{ "site": "https://getculpa.com", "entity": "Culpa", "category": "a local-first LLM cost, margin, and forecast ledger", "generated": "2026-08-03", "count": 171, "provenance": { "confidenceLevels": { "provider-reported": "Published by the party the figure describes, with no arithmetic from this site beyond any aggregation its source line names. Never read a single origin into this label: it covers a first-party usage export, a vendor's own list price, and any other published record. Each figure's source states which one and the date it was read.", "calculated": "Exact arithmetic, shown in full, over a real price-book rate. Inputs may be modelled, and the source says so wherever that applies.", "estimated": "Modelled. Always published as a range and labelled as such." }, "priceBook": { "note": "Published provider list prices, verified against the official pricing pages on 2026-07-02.", "verifiedOn": "2026-07-02" }, "limitations": [ "No first-party customer aggregates appear anywhere in this corpus.", "Worked examples marked illustrative use modelled token volumes priced at real published rates.", "Model pricing changes frequently. Check the effectiveDate on any page carrying rates." ] }, "endpoints": { "catalogue": "https://getculpa.com/api/pages", "pageText": "https://getculpa.com/{slug}/llms.txt", "hub": "https://getculpa.com/guides", "sitemap": "https://getculpa.com/sitemap.xml", "siteText": "https://getculpa.com/llms.txt" }, "pages": [ { "slug": "/agent-fan-out-cost", "url": "https://getculpa.com/agent-fan-out-cost", "textUrl": "https://getculpa.com/agent-fan-out-cost/llms.txt", "pageType": "problem", "intent": "FIX", "primaryKeyword": "agent fan out cost", "title": "Agent Fan Out Cost: Why One Request Becomes Thirty Calls", "answer": "Agent fan out is the gap between one user request and the many model calls an agent makes to satisfy it. Cost scales with steps, not requests, so a modest request count can carry a large bill. Culpa, a local-first LLM cost, margin, and forecast ledger, traces every step in a run back to the conversation that started it.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$0.39", "claim": "modelled cost of one twelve-step agent run on a frontier model", "confidence": "calculated", "source": "Sum of twelve steps with input growing 2,000 to 8,600 tokens, priced at Claude Opus 4.8 rates per 1k tokens from the price book, effective 2026-07-02. Step profile is modelled." }, { "value": "$0.14 to $0.39", "claim": "cost range for the same run capped at six steps versus twelve", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic above, at real Claude Opus 4.8 rates. A range because real step counts vary per run." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/agent-loop-cost", "url": "https://getculpa.com/agent-loop-cost", "textUrl": "https://getculpa.com/agent-loop-cost/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "agent loop cost", "title": "Agent Loop Cost: max_tokens Caps a Step, Not the Turn", "answer": "In a tool-use loop every request carries its own output cap, so the limit bounds a single step rather than the whole turn. Ten steps under one cap cost ten times it. Culpa, a local-first LLM cost, margin, and forecast ledger, prices every call and groups them by run, so a turn reports its own total rather than its steps.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$1.0240 against $0.1024", "claim": "modelled cost of a ten-step agent turn against the single-step cap that was supposed to bound it", "confidence": "calculated", "source": "Anthropic's thinking-steering-and-cost page, read 2026-08-05, states that in a tool-use loop each request in the turn has its own max_tokens so it doesn't bound the whole turn's spend. At the price book's real Claude Opus 5 output rate of $25.00 per million, one 4,096-token step costs $0.1024 and ten such steps cost $1.0240, exactly ten times. Output only, input excluded. The cap and step count are modelled, the rate and the behaviour are published." } ], "sources": [ { "label": "Anthropic, steering thinking and its cost", "url": "https://platform.claude.com/docs/en/build-with-claude/thinking-steering-and-cost" }, { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/ai-cogs-calculator", "url": "https://getculpa.com/ai-cogs-calculator", "textUrl": "https://getculpa.com/ai-cogs-calculator/llms.txt", "pageType": "margin", "intent": "EVALUATE", "primaryKeyword": "ai cogs calculator", "title": "AI COGS Calculator: What Serving One Customer Really Costs", "answer": "AI cost of goods sold is the model spend required to serve a paying customer, including retries and discarded output, excluding research and evaluation runs. Most teams under-count it by leaving waste out. Culpa, a local-first LLM cost, margin, and forecast ledger, meters every call including failures, so your COGS reflects spend rather than intent.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": "margin", "costFigures": [ { "value": "$700 to $880", "claim": "modelled customer COGS range, naive count versus corrected, against a $1,000 month", "confidence": "estimated", "source": "Modelled split of a $1,000 month priced at real GPT-5.4 rates from the price book, effective 2026-07-02. Published as a scenario because no customer data backs it." }, { "value": "$140 to $320", "claim": "modelled misattribution range hidden by two offsetting errors, one leg alone versus both", "confidence": "estimated", "source": "The $180 excluded retries plus the $140 wrongly included eval spend, from the teardown arithmetic." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/ai-cost-in-a-board-pack", "url": "https://getculpa.com/ai-cost-in-a-board-pack", "textUrl": "https://getculpa.com/ai-cost-in-a-board-pack/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "ai cost in a board pack", "title": "AI Cost in a Board Pack: Report the Margin, Not the Bill", "answer": "A board cares whether AI spend changes unit economics, not what it totalled. Report gross margin, cost per unit sold and the trend in both, with the bill as supporting detail. Culpa, a local-first LLM cost, margin, and forecast ledger, holds cost and revenue together so margin is a query rather than a spreadsheet.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "76.0% falling to 66.4%", "claim": "modelled gross margin before and after a 40% rise in model spend at flat revenue", "confidence": "calculated", "source": "A modelled $50,000 of monthly revenue with model spend rising from $12,000 to $16,800 in cost of goods sold. ($50,000 - $12,000) / $50,000 = 76.0% and ($50,000 - $16,800) / $50,000 = 66.4%, a fall of 9.6 percentage points. Every figure is modelled, and the comparison rather than the amounts is the point." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/ai-cost-in-investor-diligence", "url": "https://getculpa.com/ai-cost-in-investor-diligence", "textUrl": "https://getculpa.com/ai-cost-in-investor-diligence/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "ai cost in investor diligence", "title": "AI Cost in Investor Diligence: They Ask for Concentration", "answer": "Diligence tests whether your unit economics survive growth, which is a different question from whether this month looked fine. The answers need cost joined to customers, and concentration is the one most founders can't produce. Culpa, a local-first LLM cost, margin, and forecast ledger, holds cost and revenue together so those answers exist before they're asked.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "13.5x", "claim": "modelled cost per customer in the top decile against everyone else", "confidence": "calculated", "source": "A modelled 1,000 customers and $50,000 of monthly model spend with 60% concentrated in the top 100, which is $30,000 across 100 customers, or $300.00 each, against $20,000 across 900, or $22.22 each, a ratio of 13.5. Every figure is modelled, and the ratio rather than the amounts is the point." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/ai-saas-gross-margin-calculator", "url": "https://getculpa.com/ai-saas-gross-margin-calculator", "textUrl": "https://getculpa.com/ai-saas-gross-margin-calculator/llms.txt", "pageType": "margin", "intent": "EVALUATE", "primaryKeyword": "ai saas gross margin calculator", "title": "AI SaaS Gross Margin Calculator: The Four Inputs It Needs", "answer": "An AI SaaS gross margin calculator needs four inputs: revenue for the period, model spend, the share of that spend serving customers, and infrastructure that scales with usage. Most teams have the first and guess the rest. Culpa, a local-first LLM cost, margin, and forecast ledger, reconciles metered spend against revenue you import, so the middle two stop being estimates.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": "margin", "costFigures": [ { "value": "27.9% to 79.3%", "claim": "modelled gross margin spread across three plan tiers in the same month", "confidence": "estimated", "source": "Each tier computed at real GPT-5.4 rates per million from the price book, effective 2026-07-02. A range because customer counts and token volumes are modelled." }, { "value": "$10,905", "claim": "modelled total monthly model spend across all three tiers", "confidence": "calculated", "source": "$2,400 + $4,185 + $4,320 from the teardown arithmetic at real GPT-5.4 rates." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/ai-usage-cap-calculator", "url": "https://getculpa.com/ai-usage-cap-calculator", "textUrl": "https://getculpa.com/ai-usage-cap-calculator/llms.txt", "pageType": "margin", "intent": "EVALUATE", "primaryKeyword": "ai usage cap calculator", "title": "AI Usage Cap Calculator: Where to Set the Ceiling", "answer": "Set a usage cap from your 90th-percentile account so it binds the tail and almost nobody else. A cap set at the average binds half your base and protects against nothing. Culpa, a local-first LLM cost, margin, and forecast ledger, reports the spend distribution per customer, so the ceiling comes from measurement rather than instinct.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$3.20 to $160.00", "claim": "modelled monthly spend from median to heaviest account across 500 accounts", "confidence": "estimated", "source": "Both endpoints at real Gemini 2.5 Flash rates per 1k tokens from the price book, effective 2026-07-02. A range because the distribution is modelled." }, { "value": "$21.50", "claim": "modelled 90th-percentile monthly account spend, the level a cap should sit at", "confidence": "calculated", "source": "(30 x $0.30) + (5 x $2.50) per million, using real Gemini 2.5 Flash rates. Token volumes are modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/annual-ai-budget", "url": "https://getculpa.com/annual-ai-budget", "textUrl": "https://getculpa.com/annual-ai-budget/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "annual ai budget", "title": "Annual AI Budget: the Number You Commit To Crosses a Price Change", "answer": "An annual AI budget multiplies a run rate by twelve, and that arithmetic assumes both your usage and your rates hold still. Neither does, and rate changes arrive on published dates you can look up. Culpa, a local-first LLM cost, margin, and forecast ledger, forecasts from your own priced history against a price book versioned by date.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$120,000, or 33.3%", "claim": "modelled annual shortfall from budgeting at a run rate across a published rate change", "confidence": "calculated", "source": "10,000M input and 1,000M output tokens a month on Claude Sonnet 5 costs $20,000 + $10,000 = $30,000 at the introductory rates of $2.00 and $10.00 per million, and $30,000 + $15,000 = $45,000 at the $3.00 and $15.00 rates the price book carries effective 2026-09-01. Twelve months at the run rate is $360,000, against (4 x $30,000) + (8 x $45,000) = $480,000 built month by month, a shortfall of $120,000 or 33.3%. Both rates and the effective date are published. Token volumes and the budget year are modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/anthropic-api-cost", "url": "https://getculpa.com/anthropic-api-cost", "textUrl": "https://getculpa.com/anthropic-api-cost/llms.txt", "pageType": "provider_cost", "intent": "EVALUATE", "primaryKeyword": "anthropic api cost", "title": "Anthropic API Cost: Every Model Bills Output at Exactly 5x", "answer": "Anthropic API cost runs from $1.00 per million input tokens on Haiku to $10.00 on Fable, and every single model bills output at exactly five times input. Anthropic also charges to write a cache, not just to read one. Culpa, a local-first LLM cost, margin, and forecast ledger, meters cached, uncached and cache-write tokens separately per call.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "5.0x", "claim": "output-to-input ratio, identical on all ten Anthropic models", "confidence": "calculated", "source": "Output divided by input on every anthropic row in the price book, effective 2026-07-02. All ten return exactly 5." }, { "value": "1 read", "claim": "break-even point before a five-minute cache write pays for itself on Sonnet 5", "confidence": "calculated", "source": "Incremental write cost $0.0025 divided by the $0.009 per-read saving = 0.28, so the first read clears it. Real Sonnet 5 introductory rates. Matches Anthropic's published statement that a five-minute cache pays off after one read and a one-hour cache after two." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/anthropic-bill-spike", "url": "https://getculpa.com/anthropic-bill-spike", "textUrl": "https://getculpa.com/anthropic-bill-spike/llms.txt", "pageType": "problem", "intent": "AUDIT", "primaryKeyword": "anthropic bill spike", "title": "Anthropic Bill Spike: Cache Writes Are Usually the Cause", "answer": "An Anthropic bill spike on flat traffic usually traces to prompt caching, because Anthropic charges to create a cache and not only to read one. A five-minute write bills at 1.25 times the input rate, so a prefix you rebuild on every call costs 25% more than never caching. Culpa, a local-first LLM cost, margin, and forecast ledger, prices cache writes, cache reads and plain input as three separate things.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$0.4875 to $3.9375", "claim": "modelled cost of 21 calls over one 30k prefix, cached well against cached badly on Claude Opus 4.8", "confidence": "estimated", "source": "Both endpoints at real Claude Opus 4.8 input, cached-read and five-minute cache-write rates from the price book, effective 2026-07-02. A range because the traffic pattern is modelled." }, { "value": "25%", "claim": "premium paid over not caching at all when every call rewrites the cache", "confidence": "calculated", "source": "The five-minute cache-write rate of $0.00625 per 1k divided by the $0.005 input rate, which is the published 1.25x multiplier." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/batch-processing-cost-optimization", "url": "https://getculpa.com/batch-processing-cost-optimization", "textUrl": "https://getculpa.com/batch-processing-cost-optimization/llms.txt", "pageType": "optimization", "intent": "IMPROVE", "primaryKeyword": "batch processing cost optimization", "title": "Batch Processing Cost Optimization: 50%, and How to Lose It", "answer": "OpenAI, Anthropic and Gemini each bill batched requests at half their standard rates in exchange for a turnaround window rather than an immediate answer. Caching stacks on top of that, though hits inside a batch are best-effort. Culpa, a local-first LLM cost, margin, and forecast ledger, measures which of your traffic could wait and what the cached share actually was.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "50%", "claim": "published batch discount against standard rates at OpenAI, Anthropic and Gemini", "confidence": "provider-reported", "source": "Anthropic states all batch usage is charged at 50% of standard API prices. OpenAI's pricing table carries a batch row at exactly half of each standard row and its Batch API guide quotes 50% lower costs with a 24-hour turnaround. Gemini lists a 50% cost reduction on its paid tier. All checked 2026-08-02." }, { "value": "$174.50 to $600", "claim": "modelled monthly cost of one eval job, standard rates down to batch with cache hits landing", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at real Claude Sonnet 5 rates from the price book, effective 2026-07-02. A range because the volumes and the cache hit outcome are modelled." } ], "sources": [ { "label": "Anthropic Message Batches", "url": "https://platform.claude.com/docs/en/build-with-claude/batch-processing" }, { "label": "OpenAI Batch API guide", "url": "https://developers.openai.com/api/docs/guides/batch" }, { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/claude-fable-5-pricing", "url": "https://getculpa.com/claude-fable-5-pricing", "textUrl": "https://getculpa.com/claude-fable-5-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "claude fable 5 pricing", "title": "Claude Fable 5 Pricing: $10 In, $50 Out, Same as Mythos 5", "answer": "Claude Fable 5 pricing is $10.00 per million input tokens and $50.00 per million output tokens, and Claude Mythos 5 bills identically at the same rate point. Cache writes cost $12.50, above the input rate. Culpa, a local-first LLM cost, margin, and forecast ledger, meters cached, uncached and cache-write tokens separately per call.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$10.00 / $50.00 per million", "claim": "Claude Fable 5 and Claude Mythos 5 input and output rates, identical on both", "confidence": "calculated", "source": "Culpa price book rows for anthropic/claude-fable-5 and anthropic/claude-mythos-5, both $0.010 and $0.050 per 1k, verified against Anthropic pricing on 2026-08-01." }, { "value": "$12.50 per million", "claim": "Fable 5 five-minute cache write rate, above the input rate", "confidence": "calculated", "source": "$0.010 per 1k input x 1.25, the published Anthropic cache-write multiplier, shown on the pricing page as $12.50 per MTok." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/claude-haiku-4-5-pricing", "url": "https://getculpa.com/claude-haiku-4-5-pricing", "textUrl": "https://getculpa.com/claude-haiku-4-5-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "claude haiku 4.5 pricing", "title": "Claude Haiku 4.5 Pricing: $1 In, $5 Out per Million Tokens", "answer": "Claude Haiku 4.5 pricing is $1.00 per million input tokens and $5.00 per million output tokens, with cached reads at $0.10 per million. It runs at call volumes where a small per-call regression becomes a large monthly number. Culpa, a local-first LLM cost, margin, and forecast ledger, meters every call so a high-volume tier can't drift unnoticed.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$1.00 / $5.00 per million", "claim": "Claude Haiku 4.5 input and output rates", "confidence": "calculated", "source": "Culpa price book row for anthropic/claude-haiku-4-5, $0.001 and $0.005 per 1k, effective 2026-07-02, verified against Anthropic pricing." }, { "value": "10x", "claim": "cost of uncached input against the cached rate on Haiku 4.5", "confidence": "calculated", "source": "$1.00 input divided by the $0.10 cached rate, from the same price-book row." }, { "value": "$1,520 to $4,400", "claim": "modelled monthly cost of a one-million-call classification step, cached versus uncached", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at real Haiku 4.5 rates. A range because call volume and prompt size are modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/claude-opus-4-8-pricing", "url": "https://getculpa.com/claude-opus-4-8-pricing", "textUrl": "https://getculpa.com/claude-opus-4-8-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "claude opus 4.8 pricing", "title": "Claude Opus 4.8 Pricing: $5 In, $25 Out, and a Tokenizer Catch", "answer": "Claude Opus 4.8 pricing is $5.00 per million input tokens and $25.00 per million output tokens, matching Opus 4.5 through Opus 5. The rate is identical and the bill isn't, because Claude 4.7 and later use a newer tokenizer. Culpa, a local-first LLM cost, margin, and forecast ledger, meters the tokens each version actually produced rather than the rate it published.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$5.00 / $25.00 per million", "claim": "Claude Opus 4.5, 4.6, 4.7, 4.8 and Opus 5 input and output rates, identical across all five", "confidence": "calculated", "source": "Culpa price book rows for anthropic/claude-opus-4-5 through 4-8 and claude-opus-5, all $0.005 and $0.025 per 1k, verified against Anthropic pricing on 2026-08-01." }, { "value": "$6.25 per million", "claim": "Opus five-minute cache write rate, above the input rate", "confidence": "calculated", "source": "$0.005 per 1k input x 1.25, the published Anthropic cache-write multiplier, mirrored from the ledger's own derivation in migration 0020." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/claude-sonnet-5-pricing", "url": "https://getculpa.com/claude-sonnet-5-pricing", "textUrl": "https://getculpa.com/claude-sonnet-5-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "claude sonnet 5 pricing", "title": "Claude Sonnet 5 Pricing: $2 In, $10 Out Until 2026-08-31", "answer": "Claude Sonnet 5 pricing is $2.00 per million input tokens and $10.00 per million output tokens on an introductory rate. From 2026-09-01 it rises to $3.00 and $15.00, a 50% increase on both sides. Culpa, a local-first LLM cost, margin, and forecast ledger, holds dated rates, so a forecast built today prices September correctly.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$2.00 / $10.00 per million", "claim": "Claude Sonnet 5 introductory input and output rates, effective 2026-07-02", "confidence": "calculated", "source": "Culpa price book row for anthropic/claude-sonnet-5, $0.002 and $0.010 per 1k, effective 2026-07-02, verified against Anthropic pricing." }, { "value": "$3.00 / $15.00 per million", "claim": "Claude Sonnet 5 standard rates, effective 2026-09-01", "confidence": "calculated", "source": "Second price book row for anthropic/claude-sonnet-5, $0.003 and $0.015 per 1k, effective 2026-09-01." }, { "value": "50%", "claim": "rate increase on both input and output at the 2026-09-01 boundary", "confidence": "calculated", "source": "($3.00 - $2.00) / $2.00, and ($15.00 - $10.00) / $10.00, from the two dated rows above." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/context-creep-cost", "url": "https://getculpa.com/context-creep-cost", "textUrl": "https://getculpa.com/context-creep-cost/llms.txt", "pageType": "problem", "intent": "FIX", "primaryKeyword": "context creep cost", "title": "Context Creep Cost: When Your Prompt Grows and Nobody Notices", "answer": "Context creep is the slow growth of your prompt across releases, as instructions, examples and history get added and nothing is removed. It raises cost per call with no traffic change at all. Culpa, a local-first LLM cost, margin, and forecast ledger, tracks tokens per call over time and maps a rise to the release that caused it.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$7.50 per token per month", "claim": "cost of one extra system-prompt token at three million calls a month", "confidence": "calculated", "source": "3,000,000 calls x $0.0025 per 1k input tokens / 1,000, using the GPT-5.4 rate from the price book, effective 2026-07-02." }, { "value": "$8,250 to $15,000", "claim": "modelled monthly input spend before and after a 900-token prompt increase", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at the real GPT-5.4 input rate. A range because the call volume is modelled." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/context-overflow-cost", "url": "https://getculpa.com/context-overflow-cost", "textUrl": "https://getculpa.com/context-overflow-cost/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "context overflow cost", "title": "Context Overflow Cost: Summarising Pays After 1.25 Calls", "answer": "A conversation that fills its window has to lose something, and the two options cost differently. Truncating is free and loses information. Summarising costs one call and cuts every later one. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each call so the break-even is measured rather than guessed.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "1.25 calls", "claim": "modelled break-even for summarising a 100,000-token conversation rather than truncating it", "confidence": "calculated", "source": "A modelled 100,000-token conversation compressed to 20,000 on Claude Haiku 4.5 at real rates of $1.00 and $5.00 per million from the price book effective 2026-07-02 costs $0.1000 + $0.1000 = $0.2000. The 80,000 tokens saved on each later call, served by Claude Sonnet 5 at its real introductory $2.00 per million, is worth $0.1600 a call. $0.2000 / $0.1600 = 1.25 calls. Rates are published, the token counts and compression ratio are modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/context-window-cost-optimization", "url": "https://getculpa.com/context-window-cost-optimization", "textUrl": "https://getculpa.com/context-window-cost-optimization/llms.txt", "pageType": "optimization", "intent": "IMPROVE", "primaryKeyword": "context window cost optimization", "title": "Context Window Cost Optimization: 200k Is a Cliff, Not a Slope", "answer": "Gemini charges $1.25 per million input on prompts up to 200k tokens and $2.50 above it, and the output rate moves with the prompt too, from $10.00 to $15.00. So one token past the boundary reprices the entire call rather than the excess. Culpa, a local-first LLM cost, margin, and forecast ledger, measures how many of your prompts sit just over it.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "98%", "claim": "modelled monthly cost increase from raising prompts one percent across the 200k boundary", "confidence": "calculated", "source": "$53,250 divided by $26,875, from the teardown arithmetic at Gemini 2.5 Pro's two published tiers." }, { "value": "$26,875 to $53,250", "claim": "modelled monthly cost of the same workload either side of the boundary", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at rates read off Gemini's pricing page 2026-08-02. The above-200k tier isn't in Culpa's price book. A range because the call volumes are modelled." } ], "sources": [ { "label": "Gemini API pricing", "url": "https://ai.google.dev/gemini-api/docs/pricing" }, { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/cost-per-support-ticket", "url": "https://getculpa.com/cost-per-support-ticket", "textUrl": "https://getculpa.com/cost-per-support-ticket/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "cost per support ticket", "title": "Cost per Support Ticket: $37 per 10,000 Is the Floor", "answer": "Anthropic publishes a worked example putting 10,000 support conversations at roughly $37.00 on Claude Haiku 4.5, about a third of a cent each. That's the floor, and escalation rate multiplies it. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each conversation so your own figure sits beside the benchmark.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "about $37.00 per 10,000 conversations", "claim": "Anthropic's published worked example for support ticket processing on Claude Haiku 4.5", "confidence": "provider-reported", "source": "Anthropic's pricing page at platform.claude.com/docs/en/docs/about-claude/pricing, read 2026-08-05, gives a worked example of about 3,700 tokens per conversation on Claude Haiku 4.5 at $1.00 and $5.00 per million, totalling about $37.00 per 10,000 tickets, which is $0.0037 per conversation. Quoted as published, with no recalculation. The escalation and retrieval adjustments discussed alongside it are modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/culpa-vs-agentwatch", "url": "https://getculpa.com/culpa-vs-agentwatch", "textUrl": "https://getculpa.com/culpa-vs-agentwatch/llms.txt", "pageType": "competitor", "intent": "COMPARE", "primaryKeyword": "culpa vs agentwatch", "title": "Culpa vs AgentWatch: Enforcement at the Edge, or a Ledger You Own", "answer": "AgentWatch is an edge gateway that detects a spiralling agent, blocks it with a 402, and keeps replay and cross-provider forensics on the traffic it routes. Culpa, a local-first LLM cost, margin, and forecast ledger, stops nothing. It runs on your own infrastructure, adds margin against revenue, and forecasts the next month.", "lastReviewed": "2026-08-03", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$39 to $99 per month", "claim": "AgentWatch's published Starter and Pro plan prices", "confidence": "provider-reported", "source": "agent-watch.dev pricing, read 2026-08-03. Starter covers 150,000 requests a month and Pro 500,000, with a free tier at 50,000." } ], "sources": [ { "label": "AgentWatch", "url": "https://agent-watch.dev" } ] }, { "slug": "/culpa-vs-arize", "url": "https://getculpa.com/culpa-vs-arize", "textUrl": "https://getculpa.com/culpa-vs-arize/llms.txt", "pageType": "competitor", "intent": "COMPARE", "primaryKeyword": "culpa vs arize", "title": "Culpa vs Arize: Spans and Gigabytes, or Dollars", "answer": "Arize AX is an agent observability and evaluation platform metered on trace spans and ingestion volume, with Phoenix as its open-source counterpart. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each provider call from a versioned price book, attributes it to a customer against revenue, and forecasts next month with a range.", "lastReviewed": "2026-08-03", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$50 per month", "claim": "Arize AX Pro's published price", "confidence": "provider-reported", "source": "arize.com/pricing, read 2026-08-03. AX Pro covers 50k trace spans a month, 10 GB ingestion volume and 30-day retention on SaaS with unlimited users. AX Free covers 25k spans, 1 GB and 15-day retention. Phoenix is published separately as open source." } ], "sources": [ { "label": "Arize", "url": "https://arize.com" }, { "label": "Arize AX pricing", "url": "https://arize.com/pricing/" } ] }, { "slug": "/culpa-vs-athina", "url": "https://getculpa.com/culpa-vs-athina", "textUrl": "https://getculpa.com/culpa-vs-athina/llms.txt", "pageType": "competitor", "intent": "COMPARE", "primaryKeyword": "culpa vs athina", "title": "Culpa vs Athina: No Published Price to Model", "answer": "Athina is a collaborative AI development platform for building, testing and monitoring AI features, with 50-plus preset evaluations. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each call from a versioned price book, attributes it to a customer against revenue, and forecasts next month with a range from your own history.", "lastReviewed": "2026-08-03", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$180.00 per month", "claim": "modelled provider cost of a nightly LLM-as-judge regression, before any user traffic", "confidence": "calculated", "source": "500 test cases x 6 evaluators x 30 nights = 90,000 judge calls, at 1,500 input and 100 output tokens each = 135M input and 9M output tokens. On Claude Haiku 4.5 at real rates of $1.00 and $5.00 per million from the price book effective 2026-07-02: $135.00 + $45.00 = $180.00. Every volume here is modelled. The rates are real." } ], "sources": [ { "label": "Athina", "url": "https://www.athina.ai" }, { "label": "Athina docs", "url": "https://docs.athina.ai" } ] }, { "slug": "/culpa-vs-braintrust", "url": "https://getculpa.com/culpa-vs-braintrust", "textUrl": "https://getculpa.com/culpa-vs-braintrust/llms.txt", "pageType": "competitor", "intent": "COMPARE", "primaryKeyword": "culpa vs braintrust", "title": "Culpa vs Braintrust: Four Meters on One Invoice", "answer": "Braintrust is an evaluation and observability platform billed on four separate meters at once. Culpa, a local-first LLM cost, margin, and forecast ledger, prices your provider calls from a versioned price book, sets them against what each customer pays, and forecasts next month. One bills you for tooling, the other measures what your product costs to run.", "lastReviewed": "2026-08-03", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$249 per month", "claim": "Braintrust's published Pro platform fee, before its three usage meters", "confidence": "provider-reported", "source": "braintrust.dev/pricing, read 2026-08-03. Pro includes $249 of model credits, 5 GB processed data then $3 per GB, 50k scores then $1.50 per 1k, and 30-day retention plus archival at $0.50 per GB per month. Starter is $0 with $10 credits, 1 GB then $4 per GB, 10k scores then $2.50 per 1k, and 14-day retention." }, { "value": "$294 per month", "claim": "the same Pro plan at a modelled 20 GB of processed data, with model spend unchanged", "confidence": "calculated", "source": "$249 platform fee plus 15 GB of overage above the 5 GB included, at the published $3 per GB: $249 + $45 = $294. The 20 GB monthly volume is modelled. The fee and the per-GB rate are published on braintrust.dev/pricing, read 2026-08-03." } ], "sources": [ { "label": "Braintrust", "url": "https://www.braintrust.dev" }, { "label": "Braintrust pricing", "url": "https://www.braintrust.dev/pricing" } ] }, { "slug": "/culpa-vs-helicone", "url": "https://getculpa.com/culpa-vs-helicone", "textUrl": "https://getculpa.com/culpa-vs-helicone/llms.txt", "pageType": "competitor", "intent": "COMPARE", "primaryKeyword": "culpa vs helicone", "title": "Culpa vs Helicone: One Is in Maintenance Mode", "answer": "Helicone is an open-source AI gateway and observability platform, acquired by Mintlify and described by its own team as running in maintenance mode. It tracks cost per session and per user on its own cloud. Culpa, a local-first LLM cost, margin, and forecast ledger, keeps that on your infrastructure and adds margin against revenue and a forecast.", "lastReviewed": "2026-08-03", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$79 to $799 per month", "claim": "Helicone's published Pro and Team plan prices, before usage-based charges", "confidence": "provider-reported", "source": "helicone.ai/pricing, read 2026-08-03. Both tiers state that usage-based pricing applies on top." } ], "sources": [ { "label": "Helicone: joining Mintlify", "url": "https://www.helicone.ai/blog/joining-mintlify" }, { "label": "Helicone pricing", "url": "https://www.helicone.ai/pricing" } ] }, { "slug": "/culpa-vs-humanloop", "url": "https://getculpa.com/culpa-vs-humanloop", "textUrl": "https://getculpa.com/culpa-vs-humanloop/llms.txt", "pageType": "competitor", "intent": "COMPARE", "primaryKeyword": "culpa vs humanloop", "title": "Culpa vs Humanloop: A Cost Model Worth Reading, Now Deleted", "answer": "Humanloop was an LLM evals platform that closed on 2025-09-08, with account data deleted that day. Its docs still describe a real cost model. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each call from a versioned price book in exact decimal and forecasts next month, so this page compares two designs rather than two products you can buy.", "lastReviewed": "2026-08-03", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$48.00", "claim": "exact decimal cost of 40M output tokens at $0.0012 per 1k, against 47.99999999999999 in float", "confidence": "calculated", "source": "Arithmetic shown in full on this page. The rate shape is drawn from the price book's cheap high-volume tier and the token volume is modelled. The float result is what IEEE-754 double multiplication returns for the same product, which is the point of the figure rather than an incidental detail. Humanloop's own OpenAPI schema at humanloop.com/docs types prompt_cost and output_cost as number, format double, read 2026-08-03." }, { "value": "53 days", "claim": "notice period between Humanloop's shutdown email and permanent deletion of account data", "confidence": "provider-reported", "source": "humanloop.com/docs/v5/guides/migrating-from-humanloop, read 2026-08-03. The guide says the notice email went to all users on 2025-07-17, billing stopped on 2025-07-30, and the platform was sunset with all data permanently deleted on 2025-09-08. 2025-07-17 to 2025-09-08 is 53 days. Reported because it's the realistic answer to how long you get, not because it was ungenerous." } ], "sources": [ { "label": "Humanloop", "url": "https://humanloop.com" }, { "label": "Migrating from Humanloop", "url": "https://humanloop.com/docs/v5/guides/migrating-from-humanloop" }, { "label": "Humanloop Prompt Log API reference", "url": "https://humanloop.com/docs/api/prompts/log" } ] }, { "slug": "/culpa-vs-langfuse", "url": "https://getculpa.com/culpa-vs-langfuse", "textUrl": "https://getculpa.com/culpa-vs-langfuse/llms.txt", "pageType": "competitor", "intent": "COMPARE", "primaryKeyword": "culpa vs langfuse", "title": "Culpa vs Langfuse: An MIT-Licensed Rival, Now Inside ClickHouse", "answer": "Langfuse is an MIT-licensed LLM engineering platform you can self-host on your own infrastructure, and it has joined ClickHouse. Culpa, a local-first LLM cost, margin, and forecast ledger, is neither more open nor more private than that. What Culpa adds is margin against the revenue a customer pays you, and a forecast of the next month.", "lastReviewed": "2026-08-03", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$29 to $199 per month", "claim": "Langfuse's published Core and Pro cloud plan prices, before usage above the included units", "confidence": "provider-reported", "source": "langfuse.com/pricing, read 2026-08-03. Both include 100k units a month and charge $8 per additional 100k. Data access is where they differ: 90 days on Core and 3 years on Pro, per the same feature table. Pro's own blurb says unlimited history, which its table contradicts, so the table is what is reported here. A Hobby tier is free at 50k units and Enterprise is listed above Pro. Self-hosting is separately available under the MIT licence." } ], "sources": [ { "label": "Langfuse", "url": "https://langfuse.com" }, { "label": "Langfuse pricing", "url": "https://langfuse.com/pricing" }, { "label": "ClickHouse acquires Langfuse", "url": "https://clickhouse.com/blog/clickhouse-acquires-langfuse-open-source-llm-observability" } ] }, { "slug": "/culpa-vs-langsmith", "url": "https://getculpa.com/culpa-vs-langsmith", "textUrl": "https://getculpa.com/culpa-vs-langsmith/llms.txt", "pageType": "competitor", "intent": "COMPARE", "primaryKeyword": "culpa vs langsmith", "title": "Culpa vs LangSmith: Both Run in Your VPC, Only One Holds Revenue", "answer": "LangSmith is LangChain's tracing and evaluation platform, and it self-hosts inside your own VPC on the Enterprise plan, so traces stay on your infrastructure. Culpa, a local-first LLM cost, margin, and forecast ledger, does that by default rather than by contract. The comparison that decides is elsewhere though: Culpa carries the revenue a customer pays you and forecasts the next month, and LangSmith claims neither.", "lastReviewed": "2026-08-03", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$80.00 per customer per month", "claim": "modelled monthly model spend for one customer, the number a tracing tool can produce and a margin question still needs more than", "confidence": "calculated", "source": "40M input and 8M output tokens a month on Claude Haiku 4.5 at real rates of $1.00 and $5.00 per million from the price book, effective 2026-07-02: (40 x $1.00) + (8 x $5.00) = $40.00 + $40.00 = $80.00. Token volumes are modelled. The point of the figure is that it settles nothing until it sits beside what that customer pays." } ], "sources": [ { "label": "LangSmith", "url": "https://www.langchain.com/langsmith" }, { "label": "Self-hosted LangSmith", "url": "https://docs.langchain.com/langsmith/self-hosted" }, { "label": "LangChain pricing", "url": "https://www.langchain.com/pricing" } ] }, { "slug": "/culpa-vs-literal-ai", "url": "https://getculpa.com/culpa-vs-literal-ai", "textUrl": "https://getculpa.com/culpa-vs-literal-ai/llms.txt", "pageType": "competitor", "intent": "COMPARE", "primaryKeyword": "culpa vs literal ai", "title": "Culpa vs Literal AI: The Price Book You Had to Maintain", "answer": "Literal AI was an LLMOps platform, discontinued with service running until 2025-10-31. It priced calls from a price book you configured and dated yourself. Culpa, a local-first LLM cost, margin, and forecast ledger, ships that price book versioned and maintained, prices every call from it, and forecasts next month from your own history.", "lastReviewed": "2026-08-03", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$25.00", "claim": "understatement on 10M input tokens when a prefix pattern prices a $5.00 model at a $2.50 rate", "confidence": "calculated", "source": "Pattern matching behaviour from docs.literalai.com/guides/settings/llm, read 2026-08-03, which gives a regular expression on the model name as the matching mechanism. Rates of $2.50 and $5.00 per million input tokens are real points from the price book effective 2026-07-02. Token volume and the mismatch scenario are modelled: 10 x $2.50 = $25.00 reported against 10 x $5.00 = $50.00 true, a $25.00 gap." } ], "sources": [ { "label": "Literal AI migration guide", "url": "https://docs.literalai.com/more/migration-guide" }, { "label": "Literal AI cost tracking settings", "url": "https://docs.literalai.com/guides/settings/llm" }, { "label": "Literal AI dashboard", "url": "https://docs.literalai.com/guides/dashboard" } ] }, { "slug": "/culpa-vs-lunary", "url": "https://getculpa.com/culpa-vs-lunary", "textUrl": "https://getculpa.com/culpa-vs-lunary/llms.txt", "pageType": "competitor", "intent": "COMPARE", "primaryKeyword": "culpa vs lunary", "title": "Culpa vs Lunary: Per Seat, Per Event, Per Month of History", "answer": "Lunary is an open-source LLM observability platform billed per seat and per event, with history length set by your plan. Culpa, a local-first LLM cost, margin, and forecast ledger, prices every call from a versioned price book, carries the revenue each customer pays you, and forecasts next month from history you keep yourself.", "lastReviewed": "2026-08-03", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$20 per user per month", "claim": "Lunary's published Team plan price, before event overages", "confidence": "provider-reported", "source": "lunary.ai/pricing, read 2026-08-03. Team includes 50k events a month then $10 per additional 50k, a maximum of 10 seats and 1 year of history. Free is $0 with 10k events, 1 seat and 1 month of history. Enterprise is custom with self-hosting, SSO, PII masking and SOC 2." }, { "value": "$200 per month", "claim": "the same Team plan for a modelled team of 10, four of whom log no events at all", "confidence": "calculated", "source": "10 seats x the published $20 per user per month = $200, against $120 for the 6 engineers alone. Team size and the split between engineering and reading seats are modelled. The per-seat rate and the 10-seat maximum are published on lunary.ai/pricing, read 2026-08-03." } ], "sources": [ { "label": "Lunary", "url": "https://lunary.ai" }, { "label": "Lunary pricing", "url": "https://lunary.ai/pricing" } ] }, { "slug": "/culpa-vs-openlit", "url": "https://getculpa.com/culpa-vs-openlit", "textUrl": "https://getculpa.com/culpa-vs-openlit/llms.txt", "pageType": "competitor", "intent": "COMPARE", "primaryKeyword": "culpa vs openlit", "title": "Culpa vs OpenLIT: Apache 2.0 and the Missing Revenue Column", "answer": "OpenLIT is an open-source agent engineering platform that self-hosts free under Apache 2.0, traces with OpenTelemetry and filters by user, session or cost. Culpa, a local-first LLM cost, margin, and forecast ledger, prices calls from a versioned price book, sets them against the revenue a customer pays, and forecasts next month.", "lastReviewed": "2026-08-03", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$200.00", "claim": "monthly under-report on modelled traffic if a rate table misses Claude Sonnet 5's 2026-09-01 rise", "confidence": "calculated", "source": "Both rate rows are real and dated in the price book: $2.00 and $10.00 per million through 2026-08-31, $3.00 and $15.00 from 2026-09-01. 100M input and 20M output tokens gives $400.00 on the first and $600.00 on the second, a $200.00 gap. Token volumes are modelled. The point is the dated rate change, which is published by the provider." } ], "sources": [ { "label": "OpenLIT", "url": "https://openlit.io" }, { "label": "OpenLIT docs", "url": "https://docs.openlit.io" } ] }, { "slug": "/culpa-vs-polarity", "url": "https://getculpa.com/culpa-vs-polarity", "textUrl": "https://getculpa.com/culpa-vs-polarity/llms.txt", "pageType": "competitor", "intent": "COMPARE", "primaryKeyword": "culpa vs polarity", "title": "Culpa vs Polarity: One Changes the Bill, One Reads It", "answer": "Polarity post-trains a compact model inside your own infrastructure and routes routine work to it, keeping a frontier model for the hard cases. Culpa, a local-first LLM cost, margin, and forecast ledger, measures what that routing is worth by pricing every call before and after. One changes the bill and the other reads it, so they compete less than the shared vocabulary suggests.", "lastReviewed": "2026-08-03", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "72% lower inference cost", "claim": "a customer result Polarity publishes on its own site, not a Culpa measurement", "confidence": "provider-reported", "source": "polarity.so, read 2026-08-03, where an unattributed stat block reads 72% lower inference cost, 2.4x faster responses and +8% accuracy over Opus 4.7. A separate testimonial from Anton Reza, CTO, says the bill was cut by nearly three quarters and accuracy held. The site names no company, so whose traffic produced the figures goes unstated." } ], "sources": [ { "label": "Polarity", "url": "https://www.polarity.so/" } ] }, { "slug": "/culpa-vs-portkey", "url": "https://getculpa.com/culpa-vs-portkey", "textUrl": "https://getculpa.com/culpa-vs-portkey/llms.txt", "pageType": "competitor", "intent": "COMPARE", "primaryKeyword": "culpa vs portkey", "title": "Culpa vs Portkey: A Budget Cap Isn't a Forecast", "answer": "Portkey is an AI gateway with observability, routing and granular budget limits that block spend at a ceiling you choose. Culpa, a local-first LLM cost, margin, and forecast ledger, prices every call from a versioned price book and forecasts next month from your own history. A cap and a forecast answer opposite questions, so most teams want both.", "lastReviewed": "2026-08-03", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$49 per month", "claim": "Portkey's published Production plan price, before overages", "confidence": "provider-reported", "source": "portkey.ai/pricing, read 2026-08-03. Production includes 100k recorded logs a month with $9 per additional 100k requests, 30 days log retention and 90 days metric retention. Free Forever covers 10k logs with 3 days log retention. Enterprise is custom." }, { "value": "$85 per month", "claim": "the same Production plan at a modelled 500k logs a month, once overages are included", "confidence": "calculated", "source": "$49 base plus 4 additional blocks of 100k logs at $9 each = $49 + $36 = $85, using the overage rate published on portkey.ai/pricing and read 2026-08-03. The 500k monthly log volume is modelled, not measured." } ], "sources": [ { "label": "Portkey", "url": "https://portkey.ai" }, { "label": "Portkey pricing", "url": "https://portkey.ai/pricing" } ] }, { "slug": "/culpa-vs-promptlayer", "url": "https://getculpa.com/culpa-vs-promptlayer", "textUrl": "https://getculpa.com/culpa-vs-promptlayer/llms.txt", "pageType": "competitor", "intent": "COMPARE", "primaryKeyword": "culpa vs promptlayer", "title": "Culpa vs PromptLayer: Retention Behind an Enterprise Call", "answer": "PromptLayer is a prompt management and evaluation platform billed on requests, eval cells and agent node executions, with data retention control published only on Enterprise. Culpa, a local-first LLM cost, margin, and forecast ledger, prices every call from a versioned price book and forecasts next month from history you hold yourself.", "lastReviewed": "2026-08-03", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$500 per month", "claim": "PromptLayer's published Team plan price, before per-transaction charges", "confidence": "provider-reported", "source": "promptlayer.com/pricing, read 2026-08-03. Team covers 25 users, 100k requests a month and 7.5k eval cell executions with pay-as-you-go at $0.002 per transaction. Pro is $49 a month at $0.003 per transaction. Free is $0. Data Retention Control and self-hosting appear on Enterprise only." }, { "value": "$1,200 per month", "claim": "modelled per-transaction charges for an agent at 8 logged steps per request on the Pro rate", "confidence": "calculated", "source": "50,000 requests x 8 logged steps = 400,000 transactions, x the published $0.003 Pro rate = $1,200. Request volume and steps per request are modelled. The per-transaction rates are published on promptlayer.com/pricing, read 2026-08-03." } ], "sources": [ { "label": "PromptLayer", "url": "https://www.promptlayer.com" }, { "label": "PromptLayer pricing", "url": "https://www.promptlayer.com/pricing/" } ] }, { "slug": "/culpa-vs-raindrop", "url": "https://getculpa.com/culpa-vs-raindrop", "textUrl": "https://getculpa.com/culpa-vs-raindrop/llms.txt", "pageType": "competitor", "intent": "COMPARE", "primaryKeyword": "culpa vs raindrop", "title": "Culpa vs Raindrop: Why It Broke and What It Cost", "answer": "Raindrop is agent observability that traces production runs, surfaces silent failures like loops and broken tools, and prices itself per event, where an event is one user message and the agent's reply. Culpa, a local-first LLM cost, margin, and forecast ledger, traces the money instead, and bills per tracked call rather than per conversation turn.", "lastReviewed": "2026-08-03", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$59 to $399 per month", "claim": "Raindrop's published Startup and Pro plan prices, before per-event charges", "confidence": "provider-reported", "source": "raindrop.ai, read 2026-08-03. Startup includes 1,000 events a month then $0.004 per event, Pro is $0.003 per event to 1M then $0.002. An event is defined there as a user message plus the agent's reply." }, { "value": "$280.00", "claim": "modelled monthly cost of the wasted attempts in a three-try retry loop", "confidence": "calculated", "source": "40,000 failures x 2 wasted attempts x $0.0035 per attempt, using real Claude Haiku 4.5 rates from the price book, effective 2026-07-02. The failure volume and token counts are modelled." } ], "sources": [ { "label": "Raindrop", "url": "https://www.raindrop.ai/" } ] }, { "slug": "/culpa-vs-respan", "url": "https://getculpa.com/culpa-vs-respan", "textUrl": "https://getculpa.com/culpa-vs-respan/llms.txt", "pageType": "competitor", "intent": "COMPARE", "primaryKeyword": "culpa vs respan", "title": "Culpa vs Respan: Per-Customer Budgets, Still No Revenue", "answer": "Respan, the platform formerly called Keywords AI, is a gateway and observability product that sets budgets per key, per customer or org-wide and blocks requests before spend runs away. Culpa, a local-first LLM cost, margin, and forecast ledger, prices those calls against what each customer pays you and forecasts next month with a range.", "lastReviewed": "2026-08-03", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$199 per month", "claim": "Respan's published Team plan price, billed yearly", "confidence": "provider-reported", "source": "respan.ai/pricing, read 2026-08-03. Team includes 100k logs and 10k scores with unlimited datasets, evaluators and prompts, additional logs at $8 per 100k and additional scores at $1 per 1k. Free is $0 with 100k logs, 1k scores, 5 datasets, 2 evaluators and 5 prompts. Enterprise is custom." }, { "value": "-$50.00 per month", "claim": "modelled margin on a customer capped at $300 of spend who pays $250", "confidence": "calculated", "source": "150M input tokens at $1.00 per million plus 30M output tokens at $5.00 per million on Claude Haiku 4.5 = $150.00 + $150.00 = $300.00 of model spend, against $250.00 of revenue, giving -$50.00. Rates are real, from the price book effective 2026-07-02. Token volumes, the cap and the revenue figure are modelled." } ], "sources": [ { "label": "Respan", "url": "https://www.respan.ai" }, { "label": "Respan pricing", "url": "https://www.respan.ai/pricing" } ] }, { "slug": "/culpa-vs-traceloop", "url": "https://getculpa.com/culpa-vs-traceloop", "textUrl": "https://getculpa.com/culpa-vs-traceloop/llms.txt", "pageType": "competitor", "intent": "COMPARE", "primaryKeyword": "culpa vs traceloop", "title": "Culpa vs Traceloop: Spans Measured, Spend Not", "answer": "Traceloop is an LLM reliability platform built on OpenTelemetry, and its own pricing meters spans rather than spend. Culpa, a local-first LLM cost, margin, and forecast ledger, prices every call from a versioned price book and forecasts next month. Traceloop answers whether the output was good. Culpa answers what it cost and whether that customer pays enough.", "lastReviewed": "2026-08-03", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$37.50 per month", "claim": "model spend of the traffic that exhausts Traceloop's published 50,000-span free tier, on a retrieval shape", "confidence": "calculated", "source": "Span allowance and span definition from traceloop.com/pricing, read 2026-08-03. 6 spans per request (1 model call plus 5 vector queries) gives 50,000 / 6 = 8,333 requests. At 2,000 input and 500 output tokens on Claude Haiku 4.5, real rates of $1.00 and $5.00 per million from the price book effective 2026-07-02: 8,333 x ($0.002 + $0.0025) = $37.50. Request shape and token counts are modelled. The rates and the allowance are published." }, { "value": "24 hours", "claim": "published data retention on Traceloop's Free Forever tier", "confidence": "provider-reported", "source": "traceloop.com/pricing, read 2026-08-03. The Free Forever card lists $0 a month, up to 50K spans a month, and 24 Hours Data Retention. Enterprise lists Custom Data Retention. Reported because retention is the ceiling on every backward-looking cost question, not because 24 hours is wrong for a free tier." } ], "sources": [ { "label": "Traceloop", "url": "https://www.traceloop.com" }, { "label": "Traceloop pricing", "url": "https://www.traceloop.com/pricing" }, { "label": "Traceloop joining ServiceNow", "url": "https://www.traceloop.com/blog/traceloop-is-joining-servicenow" } ] }, { "slug": "/currency-risk-on-ai-spend", "url": "https://getculpa.com/currency-risk-on-ai-spend", "textUrl": "https://getculpa.com/currency-risk-on-ai-spend/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "currency risk on ai spend", "title": "Currency Risk on AI Spend: Billed in Dollars, Earning in Euros", "answer": "Model providers bill in US dollars. If your revenue is in another currency, your cost of goods sold moves with an exchange rate you don't control, and your margin moves with it. Culpa, a local-first LLM cost, margin, and forecast ledger, holds cost and revenue together so the effect shows up as margin rather than as a surprise.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "6.06 percentage points", "claim": "modelled gross margin lost to a currency move alone, with usage and prices unchanged", "confidence": "calculated", "source": "A modelled 100,000 euros of monthly revenue against 30,000 dollars of model spend in cost of goods sold. At a modelled 1.10 USD per EUR the cost converts to 27,273 euros for a 72.73% gross margin, and at 0.90 it converts to 33,333 euros for 66.67%, a fall of 6.06 points. Every figure including all three rates is modelled, and no rate is presented as current." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/deepseek-deepseek-v4-flash-pricing", "url": "https://getculpa.com/deepseek-deepseek-v4-flash-pricing", "textUrl": "https://getculpa.com/deepseek-deepseek-v4-flash-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "deepseek v4 flash pricing openrouter", "title": "DeepSeek V4 Flash Pricing on OpenRouter: A Rate That Moved", "answer": "DeepSeek V4 Flash bills $0.14 per million input tokens and $0.28 output through OpenRouter, with cached input at $0.028. A routed rate carries a risk a direct one doesn't, because your invoice shows a router total rather than a line per model. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each routed call from a dated price book so a change shows up.", "lastReviewed": "2026-08-03", "effectiveDate": "2026-08-03", "calculatorMode": null, "costFigures": [ { "value": "$0.14 per million", "claim": "DeepSeek V4 Flash input rate through OpenRouter", "confidence": "calculated", "source": "openrouter.ai/api/v1/models reports 0.00000014 per input token, read 2026-08-03. Multiplied to a per-million figure. Seeded as migration 0034." }, { "value": "$57.20 to $89.60", "claim": "modelled monthly cost of one workload at the superseded rate and at the current one", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at the two real dated rates. A range because the token volumes are modelled." } ], "sources": [ { "label": "OpenRouter models", "url": "https://openrouter.ai/models" } ] }, { "slug": "/failed-call-cost", "url": "https://getculpa.com/failed-call-cost", "textUrl": "https://getculpa.com/failed-call-cost/llms.txt", "pageType": "problem", "intent": "FIX", "primaryKeyword": "failed call cost", "title": "Failed Call Cost: You Pay for Output You Never Read", "answer": "A failed call still bills for every token the model generated before your code gave up. Timeouts, cancelled streams, truncated structured output and guardrail rejections all produce tokens you pay for and never read. Culpa, a local-first LLM cost, margin, and forecast ledger, meters failed calls alongside successful ones, so the cost of failure stops hiding inside the gap between your numbers and the invoice.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$0.072", "claim": "modelled cost of one abandoned call on GPT-5.5 that delivered nothing", "confidence": "calculated", "source": "(6 x $0.005) + (1.4 x $0.030) per 1k, using real GPT-5.5 rates from the price book, effective 2026-07-02. Token counts and the abandonment point are modelled." }, { "value": "86%", "claim": "share of a successful call's cost still paid when the call is abandoned near the end", "confidence": "calculated", "source": "$0.072 divided by $0.084, from the teardown arithmetic." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/fine-tuning-vs-base-model-cost", "url": "https://getculpa.com/fine-tuning-vs-base-model-cost", "textUrl": "https://getculpa.com/fine-tuning-vs-base-model-cost/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "fine tuning vs base model cost", "title": "Fine-Tuning Cost vs Base Model: the 6:1 Crossover", "answer": "Fine-tuning cost has three parts: training, inference on the tuned model, and the frontier model you stopped using. OpenAI is winding its platform down and lists prior-generation models only, so the third part is real. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each call against your own traffic.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$3.00 and $12.00 per million", "claim": "fine-tuned gpt-4.1-2025-04-14 inference rates at standard tier, against base gpt-5.4 at $2.50 and $15.00", "confidence": "provider-reported", "source": "Read off OpenAI's pricing page at developers.openai.com/api/docs/pricing on 2026-08-05, from the fine-tuning section. The model is the dated snapshot gpt-4.1-2025-04-14 at the standard service tier, outside Batch and without the data-sharing inference discount, each of which carries different rates. The same page states the fine-tuning platform is winding down and closed to new users. This site's price book carries no gpt-4.x row of any kind, so these fine-tuned rates sit outside it and stand on the provider's page alone. The base gpt-5.4 figures of $2.50 and $15.00 on the same page do match the price book's row effective 2026-07-02 exactly. Nothing here is aggregated or recalculated." }, { "value": "6 to 1", "claim": "input to output token ratio at which fine-tuned gpt-4.1-2025-04-14 and base gpt-5.4 cost the same", "confidence": "calculated", "source": "Setting 3.00i + 12.00o equal to 2.50i + 15.00o per million gives 0.50i = 3.00o, so i = 6o. Checked at three points: at 2:1 the tuned model costs $18.00 against $20.00, at 6:1 both cost $30.00, at 10:1 the tuned model costs $42.00 against $40.00. Rates are provider-published, the ratios are modelled." } ], "sources": [ { "label": "OpenAI API pricing, including fine-tuning", "url": "https://developers.openai.com/api/docs/pricing" }, { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/free-tier-abuse", "url": "https://getculpa.com/free-tier-abuse", "textUrl": "https://getculpa.com/free-tier-abuse/llms.txt", "pageType": "problem", "intent": "FIX", "primaryKeyword": "free tier abuse", "title": "Free Tier Abuse: What One Automated Signup Really Costs", "answer": "Free tier abuse is scripted signups consuming a whole free allowance at the highest cost per call they can reach. A genuine trial user takes a fraction of what's granted, while a scripted one takes all of it with the longest prompts it can send. Culpa, a local-first LLM cost, margin, and forecast ledger, meters spend per signup, so the free tier carries a unit cost instead of a hope.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$0.0495 to $4.95", "claim": "modelled cost of one free signup, genuine trial against scripted, on GPT-5.4 mini", "confidence": "estimated", "source": "Both endpoints at real GPT-5.4 mini rates per 1k tokens from the price book, effective 2026-07-02. A range because the usage profiles are modelled." }, { "value": "81%", "claim": "modelled share of free-tier spend produced by 4% of signups", "confidence": "calculated", "source": "$1,782.00 divided by $2,209.68, from the teardown arithmetic." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/free-trial-llm-cost", "url": "https://getculpa.com/free-trial-llm-cost", "textUrl": "https://getculpa.com/free-trial-llm-cost/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "free trial llm cost", "title": "Free Trial LLM Cost: Compute Is Customer Acquisition Cost", "answer": "Free trial LLM cost is the model spend consumed by users who haven't paid, which makes it customer acquisition cost that arrives on a provider invoice rather than in an ad account. Divide it by your conversion rate for the real figure. Culpa, a local-first LLM cost, margin, and forecast ledger, prices trial calls and attributes them per account.", "lastReviewed": "2026-08-03", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$52.50", "claim": "modelled trial compute per acquired customer, against $4.20 per trial", "confidence": "calculated", "source": "1,000 modelled trials each consuming 3M input and 240k output tokens on Claude Haiku 4.5 at real rates of $1.00 and $5.00 per million from the price book effective 2026-07-02 gives $4.20 per trial and $4,200.00 a month. At a modelled 8% conversion that's 80 customers, so $4,200.00 / 80 = $52.50 each. Token volumes, trial count and conversion rate are all modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/gemini-2-5-flash-pricing", "url": "https://getculpa.com/gemini-2-5-flash-pricing", "textUrl": "https://getculpa.com/gemini-2-5-flash-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "gemini 2.5 flash pricing", "title": "Gemini 2.5 Flash Pricing: $0.30 In, $2.50 Out per Million", "answer": "Gemini 2.5 Flash pricing is $0.30 per million input tokens and $2.50 per million output tokens, a spread of roughly eight to one. That makes answer length, not prompt length, the lever that moves your bill. Culpa, a local-first LLM cost, margin, and forecast ledger, splits input and output per call so you can see which side is growing.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$0.30 / $2.50 per million", "claim": "Gemini 2.5 Flash input and output rates", "confidence": "calculated", "source": "Culpa price book row for gemini/gemini-2.5-flash, $0.0003 and $0.0025 per 1k, effective 2026-07-02, verified against Gemini API pricing." }, { "value": "8.3x", "claim": "output rate as a multiple of the input rate on Gemini 2.5 Flash", "confidence": "calculated", "source": "$2.50 divided by $0.30, from the same price-book row." } ], "sources": [ { "label": "Gemini API pricing", "url": "https://ai.google.dev/gemini-api/docs/pricing" } ] }, { "slug": "/gemini-2-5-flash-vs-llama-3-3-70b-versatile-cost", "url": "https://getculpa.com/gemini-2-5-flash-vs-llama-3-3-70b-versatile-cost", "textUrl": "https://getculpa.com/gemini-2-5-flash-vs-llama-3-3-70b-versatile-cost/llms.txt", "pageType": "model_compare", "intent": "COMPARE", "primaryKeyword": "gemini 2.5 flash vs llama 3.3 70b cost", "title": "Gemini 2.5 Flash vs Llama 3.3 70B Cost: Where They Cross", "answer": "Gemini 2.5 Flash costs less on input and far more on output than Llama 3.3 70B on Groq, so they cross at an output-to-input ratio of about 0.17. Retrieval workloads sit below that line and chat workloads above it. Culpa, a local-first LLM cost, margin, and forecast ledger, prices both against your own split so the choice stops being a preference.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "0.17", "claim": "output-to-input ratio at which these two models cost the same", "confidence": "calculated", "source": "($0.30 - $0.59) divided by ($0.79 - $2.50), using real Gemini 2.5 Flash and Groq Llama 3.3 70B rates per million from the price book, effective 2026-07-02." }, { "value": "$90.60 to $251.80", "claim": "modelled monthly cost across two workloads and two models, cheapest to dearest", "confidence": "estimated", "source": "The four totals in the teardown arithmetic at real rates. A range because the token volumes are modelled." } ], "sources": [ { "label": "Gemini API pricing", "url": "https://ai.google.dev/gemini-api/docs/pricing" }, { "label": "Groq pricing", "url": "https://groq.com/pricing" } ] }, { "slug": "/gemini-2-5-pro-pricing", "url": "https://getculpa.com/gemini-2-5-pro-pricing", "textUrl": "https://getculpa.com/gemini-2-5-pro-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "gemini 2.5 pro pricing", "title": "Gemini 2.5 Pro Pricing: $1.25 In, $10.00 Out per Million", "answer": "Gemini 2.5 Pro pricing is $1.25 per million input tokens and $10.00 per million output tokens, an eight-to-one spread. Cheap input makes long-context analysis viable, and expensive output punishes narration. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each call from that rate and attributes it to the document run that spent it.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$1.25 / $10.00 per million", "claim": "Gemini 2.5 Pro input and output rates", "confidence": "calculated", "source": "Culpa price book row for gemini/gemini-2.5-pro, $0.00125 and $0.010 per 1k, effective 2026-07-02, verified against Gemini API pricing." }, { "value": "$260 to $460", "claim": "modelled cost of one 4,000-document batch run, at halved versus full context", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at real Gemini 2.5 Pro rates. A range because document count and context size are modelled." } ], "sources": [ { "label": "Gemini API pricing", "url": "https://ai.google.dev/gemini-api/docs/pricing" } ] }, { "slug": "/gemini-2-5-pro-vs-gemini-3-5-flash-cost", "url": "https://getculpa.com/gemini-2-5-pro-vs-gemini-3-5-flash-cost", "textUrl": "https://getculpa.com/gemini-2-5-pro-vs-gemini-3-5-flash-cost/llms.txt", "pageType": "model_compare", "intent": "COMPARE", "primaryKeyword": "gemini 2.5 pro vs gemini 3.5 flash cost", "title": "Gemini 2.5 Pro vs Gemini 3.5 Flash Cost: The Crossover", "answer": "Gemini 2.5 Pro costs less than Gemini 3.5 Flash on input-heavy work, and more on output-heavy work. They cross at an output-to-input ratio of 0.25, so the cheaper model is decided by your own token mix rather than by the tier in the name. Culpa, a local-first LLM cost, margin, and forecast ledger, prices both against your real traffic.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "0.25", "claim": "output-to-input ratio at which these two models cost exactly the same", "confidence": "calculated", "source": "($1.25 - $1.50) divided by ($9.00 - $10.00), using real Gemini 2.5 Pro and 3.5 Flash rates per million from the price book, effective 2026-07-02." }, { "value": "$450 to $625", "claim": "modelled monthly cost of Gemini 2.5 Pro across a retrieval and a chat workload", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at real rates. A range because the token volumes are modelled." } ], "sources": [ { "label": "Gemini API pricing", "url": "https://ai.google.dev/gemini-api/docs/pricing" } ] }, { "slug": "/gemini-3-1-flash-lite-pricing", "url": "https://getculpa.com/gemini-3-1-flash-lite-pricing", "textUrl": "https://getculpa.com/gemini-3-1-flash-lite-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "gemini 3.1 flash lite pricing", "title": "Gemini 3.1 Flash-Lite Pricing: The Guardrail Tax", "answer": "Gemini 3.1 Flash-Lite pricing is $0.25 per million input tokens and $1.50 per million output tokens. Its real cost shows up in call count rather than in rate, because a model this cheap gets put in front of every request as a router or a guardrail. Culpa, a local-first LLM cost, margin, and forecast ledger, counts those pre-filter calls as a feature of their own.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$0.25 / $1.50 per million", "claim": "Gemini 3.1 Flash-Lite input and output rates", "confidence": "calculated", "source": "Culpa price book row for gemini/gemini-3.1-flash-lite, $0.00025 and $0.0015 per 1k, effective 2026-07-02, verified against Gemini API pricing. Shown here per million." }, { "value": "$1,225", "claim": "modelled monthly cost of running this model as a pre-filter on 5 million requests", "confidence": "calculated", "source": "5,000,000 x $0.000245 per call, using real Gemini 3.1 Flash-Lite rates per 1k. Request volume and token counts are modelled." }, { "value": "2.2%", "claim": "modelled break-even diversion rate for the pre-filter against a Gemini 2.5 Pro workload", "confidence": "calculated", "source": "$1,225 divided by the $0.011 per-call saving, from the teardown arithmetic at real rates for both models." } ], "sources": [ { "label": "Gemini API pricing", "url": "https://ai.google.dev/gemini-api/docs/pricing" } ] }, { "slug": "/gemini-3-1-flash-lite-vs-llama-3-3-70b-versatile-cost", "url": "https://getculpa.com/gemini-3-1-flash-lite-vs-llama-3-3-70b-versatile-cost", "textUrl": "https://getculpa.com/gemini-3-1-flash-lite-vs-llama-3-3-70b-versatile-cost/llms.txt", "pageType": "model_compare", "intent": "COMPARE", "primaryKeyword": "gemini 3.1 flash lite vs llama 3.3 70b cost", "title": "Gemini 3.1 Flash-Lite vs Llama 3.3 70B Cost: A Crossover You Reach", "answer": "Gemini 3.1 Flash-Lite costs less than Llama 3.3 70B on Groq below an output-to-input ratio of 0.48, and more above it. That crossover sits inside the range real features occupy, so one product often has work on both sides of it. Culpa, a local-first LLM cost, margin, and forecast ledger, measures the ratio per feature and prices each against both.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "0.48", "claim": "output-to-input ratio at which these two models cost the same", "confidence": "calculated", "source": "($0.59 - $0.25) divided by ($1.50 - $0.79), using real Gemini 3.1 Flash-Lite and Groq Llama 3.3 70B Versatile rates per million from the price book, effective 2026-07-02 and re-verified on both provider pages 2026-08-02." }, { "value": "13%", "claim": "modelled saving from routing two features to opposite models rather than picking the best single one", "confidence": "calculated", "source": "$197.50 minus $171.90, divided by $197.50, from the teardown arithmetic." }, { "value": "$76.90 to $141.70", "claim": "modelled monthly cost across two features and two models, cheapest to dearest", "confidence": "estimated", "source": "The four totals in the teardown arithmetic at real rates. A range because the token volumes are modelled." } ], "sources": [ { "label": "Gemini API pricing", "url": "https://ai.google.dev/gemini-api/docs/pricing" }, { "label": "Groq pricing", "url": "https://groq.com/pricing" } ] }, { "slug": "/gemini-3-1-pro-preview-pricing", "url": "https://getculpa.com/gemini-3-1-pro-preview-pricing", "textUrl": "https://getculpa.com/gemini-3-1-pro-preview-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "gemini 3.1 pro preview pricing", "title": "Gemini 3.1 Pro Preview Pricing: A Rate With an Expiry", "answer": "Gemini 3.1 Pro Preview pricing is $2.00 per million input tokens and $12.00 per million output tokens. Preview rates carry no commitment, so a forecast built on this one ages faster than any other number in your model. Culpa, a local-first LLM cost, margin, and forecast ledger, prices every call from a dated price book, so a rate change lands as a diff rather than as a surprise.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$2.00 / $12.00 per million", "claim": "Gemini 3.1 Pro Preview input and output rates, with no commitment to stay there", "confidence": "calculated", "source": "Culpa price book row for gemini/gemini-3.1-pro-preview, $0.002 and $0.012 per 1k, effective 2026-07-02, verified against Gemini API pricing. Shown here per million." }, { "value": "$650 to $1,320", "claim": "modelled monthly cost band for the same workload if the preview rate settles lower or rises", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic. The lower uses real Gemini 2.5 Pro rates and the upper applies the 50% rise Claude Sonnet 5's introductory pricing took. Volumes are modelled and neither landing point is announced." } ], "sources": [ { "label": "Gemini API pricing", "url": "https://ai.google.dev/gemini-api/docs/pricing" } ] }, { "slug": "/gemini-3-5-flash-pricing", "url": "https://getculpa.com/gemini-3-5-flash-pricing", "textUrl": "https://getculpa.com/gemini-3-5-flash-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "gemini 3.5 flash pricing", "title": "Gemini 3.5 Flash Pricing: The Newer Flash Costs 5x More", "answer": "Gemini 3.5 Flash pricing is $1.50 per million input tokens and $9.00 per million output tokens, five times the input rate of Gemini 2.5 Flash at $0.30. A version bump inside the same family multiplies your bill rather than trimming it. Culpa, a local-first LLM cost, margin, and forecast ledger, prices the same workload on both, so the move stays a decision.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$1.50 / $9.00 per million", "claim": "Gemini 3.5 Flash input and output rates", "confidence": "calculated", "source": "Culpa price book row for gemini/gemini-3.5-flash, $0.0015 and $0.009 per 1k, effective 2026-07-02, verified against Gemini API pricing. Shown here per million." }, { "value": "5x", "claim": "Gemini 3.5 Flash input rate as a multiple of Gemini 2.5 Flash", "confidence": "calculated", "source": "$1.50 divided by $0.30, from the gemini-3.5-flash and gemini-2.5-flash price-book rows." }, { "value": "$61.50 to $255.00", "claim": "modelled monthly cost of one identical workload on 2.5 Flash against 3.5 Flash", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at real rates for each model. A range because the token volumes are modelled." } ], "sources": [ { "label": "Gemini API pricing", "url": "https://ai.google.dev/gemini-api/docs/pricing" } ] }, { "slug": "/gemini-api-cost", "url": "https://getculpa.com/gemini-api-cost", "textUrl": "https://getculpa.com/gemini-api-cost/llms.txt", "pageType": "provider_cost", "intent": "EVALUATE", "primaryKeyword": "gemini api cost", "title": "Gemini API Cost: The Widest Output Spreads, Up to 8.3x", "answer": "Gemini API cost runs from $0.25 per million input tokens on Flash Lite to $2.00 on the Pro preview, with output spreads reaching 8.3 times input. That makes answer length the dominant lever here. Culpa, a local-first LLM cost, margin, and forecast ledger, splits input and output per call, so you can see which side is actually growing.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "8.3x", "claim": "widest output-to-input ratio across the Gemini line, on 2.5 Flash", "confidence": "calculated", "source": "$2.50 output divided by $0.30 input per million on gemini-2.5-flash, from the price book, effective 2026-07-02." }, { "value": "$0.55 to $4.40", "claim": "cost of one million input and 200,000 output tokens across the five Gemini models", "confidence": "calculated", "source": "Five calculations shown in the teardown, all from price-book rows effective 2026-07-02." } ], "sources": [ { "label": "Gemini API pricing", "url": "https://ai.google.dev/gemini-api/docs/pricing" } ] }, { "slug": "/gemini-bill-spike", "url": "https://getculpa.com/gemini-bill-spike", "textUrl": "https://getculpa.com/gemini-bill-spike/llms.txt", "pageType": "problem", "intent": "AUDIT", "primaryKeyword": "gemini bill spike", "title": "Gemini Bill Spike: The 200k Context Tier Nobody Priced", "answer": "A Gemini bill spike with no traffic change often means prompts crossed 200,000 tokens, where input doubles and output rises by half. Crossing that line changes your rate as well as your volume, so a modest prompt increase lands as a large bill increase. Culpa, a local-first LLM cost, margin, and forecast ledger, prices every call from a dated price book and names the feature behind the change.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$0.245", "claim": "modelled cost of one 180k-token call on Gemini 2.5 Pro below the long-context boundary", "confidence": "calculated", "source": "(180 x $0.00125) + (2 x $0.010) per 1k, using real Gemini 2.5 Pro rates from the price book, effective 2026-07-02. Token volumes are modelled." }, { "value": "$0.245 to $0.630", "claim": "modelled cost per call either side of the 200,000-token tier boundary", "confidence": "estimated", "source": "The upper endpoint uses the above-200k tier listed on the Gemini pricing page, fetched 2026-08-01, which the price book deliberately doesn't hold. A range because token volumes are modelled." } ], "sources": [ { "label": "Gemini API pricing", "url": "https://ai.google.dev/gemini-api/docs/pricing" } ] }, { "slug": "/gpt-5-4-mini-pricing", "url": "https://getculpa.com/gpt-5-4-mini-pricing", "textUrl": "https://getculpa.com/gpt-5-4-mini-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "gpt-5.4 mini pricing", "title": "GPT-5.4 mini Pricing: $0.75 In, $4.50 Out per Million", "answer": "GPT-5.4 mini pricing is $0.75 per million input tokens and $4.50 per million output tokens, roughly a third of GPT-5.4 on both sides. Cheap enough that teams stop counting calls, which is how volume eats the saving. Culpa, a local-first LLM cost, margin, and forecast ledger, meters every call so a cheap model can't hide behind its rate.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$0.75 / $4.50 per million", "claim": "GPT-5.4 mini input and output rates", "confidence": "calculated", "source": "Culpa price book row for openai/gpt-5.4-mini, $0.00075 and $0.0045 per 1k, effective 2026-07-02, verified against OpenAI API pricing. Shown here per million." }, { "value": "$4,560 to $7,560", "claim": "modelled monthly cost of a mini routing step at eight million calls, trimmed prompt versus untrimmed", "confidence": "estimated", "source": "From the teardown arithmetic at real GPT-5.4 mini rates. Published as a modelled scenario because the call volume and token sizes are assumptions." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/gpt-5-4-mini-vs-kimi-k2-instruct-0905-cost", "url": "https://getculpa.com/gpt-5-4-mini-vs-kimi-k2-instruct-0905-cost", "textUrl": "https://getculpa.com/gpt-5-4-mini-vs-kimi-k2-instruct-0905-cost/llms.txt", "pageType": "model_compare", "intent": "COMPARE", "primaryKeyword": "gpt-5.4 mini vs kimi k2 cost", "title": "GPT-5.4 Mini vs Kimi K2 Cost: The Cheaper Input Rate Loses", "answer": "GPT-5.4 mini lists an input rate a quarter below Kimi K2's and still costs more on most real traffic, because its output rate runs half again higher. The two cross at an output-to-input ratio of 0.17. Culpa, a local-first LLM cost, margin, and forecast ledger, measures your own ratio and prices both models against it.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "0.17", "claim": "output-to-input ratio at which these two models cost the same", "confidence": "calculated", "source": "($1.00 - $0.75) divided by ($4.50 - $3.00), using real Groq Kimi K2 Instruct 0905 and OpenAI GPT-5.4 mini rates per million from the price book, effective 2026-07-02 and re-verified on both provider pages 2026-08-02." }, { "value": "$120 to $255", "claim": "modelled monthly cost across two workloads and two models, cheapest to dearest", "confidence": "estimated", "source": "The four totals in the teardown arithmetic at real rates. A range because the token volumes are modelled." } ], "sources": [ { "label": "Groq pricing", "url": "https://groq.com/pricing" }, { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/gpt-5-4-nano-pricing", "url": "https://getculpa.com/gpt-5-4-nano-pricing", "textUrl": "https://getculpa.com/gpt-5-4-nano-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "gpt-5.4 nano pricing", "title": "GPT-5.4 Nano Pricing: The Bottom of a 150x Ladder", "answer": "GPT-5.4 nano pricing is $0.20 per million input tokens and $1.25 per million output tokens, which is 150 times below the Pro tier's $30.00 input on the same rate card. The widest cost decision you'll make sits inside one vendor rather than between two. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each call by the model that served it.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$0.20 / $1.25 per million", "claim": "GPT-5.4 nano input and output rates", "confidence": "calculated", "source": "Culpa price book row for openai/gpt-5.4-nano, $0.0002 and $0.00125 per 1k, effective 2026-07-02, verified against OpenAI API pricing. Shown here per million." }, { "value": "150x", "claim": "input rate of the Pro tier as a multiple of nano, on the same OpenAI rate card", "confidence": "calculated", "source": "$30.00 divided by $0.20, from the gpt-5.4-pro and gpt-5.4-nano price-book rows." }, { "value": "$22.50 to $3,300.00", "claim": "modelled monthly cost of one identical workload across four OpenAI tiers", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at real OpenAI rates. A range because the token volumes are modelled." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/gpt-5-4-nano-vs-llama-3-3-70b-versatile-cost", "url": "https://getculpa.com/gpt-5-4-nano-vs-llama-3-3-70b-versatile-cost", "textUrl": "https://getculpa.com/gpt-5-4-nano-vs-llama-3-3-70b-versatile-cost/llms.txt", "pageType": "model_compare", "intent": "COMPARE", "primaryKeyword": "gpt-5.4 nano vs llama 3.3 70b cost", "title": "GPT-5.4 Nano vs Llama 3.3 70B Cost: A Crossover at 0.85", "answer": "GPT-5.4 nano costs less than Llama 3.3 70B on Groq until your output volume reaches about 0.85 of your input volume, which almost no real workload does. Culpa, a local-first LLM cost, margin, and forecast ledger, prices both against your own split, and here the crossover sits so high that the answer is settled for most traffic.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "0.85", "claim": "output-to-input ratio at which these two models cost the same", "confidence": "calculated", "source": "($0.59 - $0.20) divided by ($1.25 - $0.79), using real GPT-5.4 nano and Groq Llama 3.3 70B rates per million from the price book, effective 2026-07-02." }, { "value": "$76.90 to $157.50", "claim": "modelled monthly cost across two workloads and two models, cheapest to dearest", "confidence": "estimated", "source": "The four totals in the teardown arithmetic at real rates. A range because the token volumes are modelled." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" }, { "label": "Groq pricing", "url": "https://groq.com/pricing" } ] }, { "slug": "/gpt-5-4-pricing", "url": "https://getculpa.com/gpt-5-4-pricing", "textUrl": "https://getculpa.com/gpt-5-4-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "gpt-5.4 pricing", "title": "GPT-5.4 Pricing: $2.50 In, $15.00 Out per Million Tokens", "answer": "GPT-5.4 pricing is $2.50 per million input tokens and $15.00 per million output tokens, on rates verified 2026-07-02. Output bills six times input, so answer length drives your bill harder than prompt length. Culpa, a local-first LLM cost, margin, and forecast ledger, prices every call from that rate and names the conversation that spent it.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$2.50 / $15.00 per million", "claim": "GPT-5.4 input and output rates", "confidence": "calculated", "source": "Culpa price book row for openai/gpt-5.4, $0.0025 and $0.015 per 1k, effective 2026-07-02, verified against OpenAI API pricing. Shown here per million." }, { "value": "6x", "claim": "output rate as a multiple of the input rate on GPT-5.4", "confidence": "calculated", "source": "$15.00 divided by $2.50, from the same price-book row." }, { "value": "90%", "claim": "discount the cached input rate carries against standard input", "confidence": "calculated", "source": "1 - ($0.25 / $2.50), using the cached and input rates from the same row." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/gpt-5-4-pro-pricing", "url": "https://getculpa.com/gpt-5-4-pro-pricing", "textUrl": "https://getculpa.com/gpt-5-4-pro-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "gpt-5.4 pro pricing", "title": "GPT-5.4 Pro Pricing: $30 In, $180 Out, and No Cached Rate", "answer": "GPT-5.4 Pro pricing is $30.00 per million input tokens and $180.00 per million output tokens, and GPT-5.5 Pro bills identically. Neither carries a cached rate, so a repeated system prompt pays full price every call. Culpa, a local-first LLM cost, margin, and forecast ledger, meters the repeated share of your input so that cost stops being invisible.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$30.00 / $180.00 per million", "claim": "GPT-5.4 Pro and GPT-5.5 Pro input and output rates, identical on both models", "confidence": "calculated", "source": "Culpa price book rows for openai/gpt-5.4-pro and openai/gpt-5.5-pro, both $0.030 and $0.180 per 1k, effective 2026-07-02, verified against OpenAI API pricing." }, { "value": "12x", "claim": "Pro tier cost as a multiple of GPT-5.4 on both input and output", "confidence": "calculated", "source": "$30.00 / $2.50 and $180.00 / $15.00, from the price-book rows." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/gpt-5-5-pricing", "url": "https://getculpa.com/gpt-5-5-pricing", "textUrl": "https://getculpa.com/gpt-5-5-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "gpt-5.5 pricing", "title": "GPT-5.5 Pricing: $5.00 In, $30.00 Out, Same as GPT-5.6 Sol", "answer": "GPT-5.5 pricing is $5.00 per million input tokens and $30.00 per million output tokens, exactly double GPT-5.4 on both sides. The premium earns itself only when a wrong answer costs more than the call. Culpa, a local-first LLM cost, margin, and forecast ledger, meters both models side by side so the comparison is measured, not argued.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$5.00 / $30.00 per million", "claim": "GPT-5.5 input and output rates", "confidence": "calculated", "source": "Culpa price book row for openai/gpt-5.5, $0.005 and $0.030 per 1k, effective 2026-07-02, verified against OpenAI API pricing. Shown here per million." }, { "value": "2.0x", "claim": "GPT-5.5 cost relative to GPT-5.4 on both input and output", "confidence": "calculated", "source": "$5.00 / $2.50 and $30.00 / $15.00, from the two price-book rows." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/gpt-5-6-luna-pricing", "url": "https://getculpa.com/gpt-5-6-luna-pricing", "textUrl": "https://getculpa.com/gpt-5-6-luna-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "gpt-5.6 luna pricing", "title": "GPT-5.6 Luna Pricing: $0.20 In, $1.20 Out per Million", "answer": "GPT-5.6 Luna pricing is $0.20 per million input tokens and $1.20 per million output tokens, the cheapest OpenAI tier and fractionally under GPT-5.4 nano on output. At this rate call volume decides your bill rather than the rate. Culpa, a local-first LLM cost, margin, and forecast ledger, meters every call so a cheap tier can't hide behind its price.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-08-01", "calculatorMode": null, "costFigures": [ { "value": "$0.20 / $1.20 per million", "claim": "GPT-5.6 Luna input and output rates", "confidence": "calculated", "source": "Culpa price book row for openai/gpt-5.6-luna, $0.0002 and $0.0012 per 1k, read off the OpenAI pricing page on 2026-08-01. Pending seed into the ledger migrations." }, { "value": "$600 to $1,400", "claim": "modelled monthly input cost at ten million calls, trimmed prompt versus untrimmed", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at the real Luna input rate. A range because call volume and prompt size are modelled." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/gpt-5-6-luna-vs-llama-3-3-70b-versatile-cost", "url": "https://getculpa.com/gpt-5-6-luna-vs-llama-3-3-70b-versatile-cost", "textUrl": "https://getculpa.com/gpt-5-6-luna-vs-llama-3-3-70b-versatile-cost/llms.txt", "pageType": "model_compare", "intent": "COMPARE", "primaryKeyword": "gpt-5.6 luna vs llama 3.3 70b cost", "title": "GPT-5.6 Luna vs Llama 3.3 70B Cost: The Crossover Moves", "answer": "GPT-5.6 luna undercuts Llama 3.3 70B on Groq for nearly every short-context workload and loses to it on nearly every long-context one. OpenAI charges a second, higher rate once a prompt gets long, which moves the crossover from 0.95 down to 0.19. Culpa, a local-first LLM cost, margin, and forecast ledger, meters which tier your calls landed in.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "0.95", "claim": "output-to-input ratio at which these two cost the same on short-context prompts", "confidence": "calculated", "source": "($0.59 - $0.20) divided by ($1.20 - $0.79), using real Groq Llama 3.3 70B Versatile and OpenAI GPT-5.6 luna short-context rates per million from the price book, effective 2026-07-02 and re-verified 2026-08-02." }, { "value": "0.19", "claim": "the same crossover once a prompt prices at OpenAI's long-context rate", "confidence": "calculated", "source": "($0.59 - $0.40) divided by ($1.80 - $0.79), using the long-context rates read off OpenAI's pricing page 2026-08-02. Those rates aren't in Culpa's price book yet." }, { "value": "$68.00 to $112.00", "claim": "modelled monthly cost of one workload across both context tiers and both models", "confidence": "estimated", "source": "The totals in the teardown arithmetic at real rates. A range because the token volumes are modelled." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" }, { "label": "Groq pricing", "url": "https://groq.com/pricing" } ] }, { "slug": "/gpt-5-6-terra-pricing", "url": "https://getculpa.com/gpt-5-6-terra-pricing", "textUrl": "https://getculpa.com/gpt-5-6-terra-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "gpt-5.6 terra pricing", "title": "GPT-5.6 Terra Pricing: $2.00 In, $12.00 Out per Million", "answer": "GPT-5.6 Terra pricing is $2.00 per million input tokens and $12.00 per million output tokens, which undercuts GPT-5.4 on both sides at $2.50 and $15.00. Output bills six times input, the usual OpenAI shape. Culpa, a local-first LLM cost, margin, and forecast ledger, records the model on every call, so a migration lands as a measured delta.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-08-01", "calculatorMode": null, "costFigures": [ { "value": "$2.00 / $12.00 per million", "claim": "GPT-5.6 Terra input and output rates", "confidence": "calculated", "source": "Culpa price book row for openai/gpt-5.6-terra, $0.002 and $0.012 per 1k, read off the OpenAI pricing page on 2026-08-01. Pending seed into the ledger migrations." }, { "value": "20%", "claim": "Terra discount against GPT-5.4 on both input and output", "confidence": "calculated", "source": "($2.50 - $2.00) / $2.50 and ($15.00 - $12.00) / $15.00, from the two price-book rows." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/granular-llm-cost-attribution", "url": "https://getculpa.com/granular-llm-cost-attribution", "textUrl": "https://getculpa.com/granular-llm-cost-attribution/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "granular llm cost attribution", "title": "Granular LLM Cost Attribution: Every Tool Has a Floor", "answer": "Granular LLM cost attribution means pricing each individual call and rolling it up to the conversation, customer or feature that caused it. Every tool has a floor below which it aggregates, and nothing recovers detail once it has gone. Culpa, a local-first LLM cost, margin, and forecast ledger, prices every call and keeps that floor at the call.", "lastReviewed": "2026-08-03", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$178.00", "claim": "modelled monthly cost of one shared agent, hiding a 9.2x spread between the three customers behind it", "confidence": "calculated", "source": "Three modelled customers priced on Claude Haiku 4.5 at real rates of $1.00 and $5.00 per million from the price book effective 2026-07-02: $120.00 + $45.00 + $13.00 = $178.00. Customer token volumes are modelled. The point of the figure is that the total is correct and the three components exist nowhere in an agent-level record." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" }, { "label": "Grafana Cloud AI observability", "url": "https://grafana.com/docs/grafana-cloud/monitor-applications/ai-observability/" } ] }, { "slug": "/groq-api-cost", "url": "https://getculpa.com/groq-api-cost", "textUrl": "https://getculpa.com/groq-api-cost/llms.txt", "pageType": "provider_cost", "intent": "EVALUATE", "primaryKeyword": "groq api cost", "title": "Groq API Cost: Where Trimming Your Output Barely Helps", "answer": "Groq API cost runs from $0.05 per million input tokens to $1.00, and output ratios go as flat as 1.34 times input on Llama 3.3 70B. That flatness breaks the usual advice to trim the answer first. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each call from the model's own rates rather than a rule of thumb.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "1.34x", "claim": "flattest output-to-input ratio in this price book, on Llama 3.3 70B via Groq", "confidence": "calculated", "source": "$0.79 output divided by $0.59 input per million, from the price book, effective 2026-07-02." }, { "value": "20x", "claim": "spread between the cheapest and dearest Groq input rates", "confidence": "calculated", "source": "$1.00 per million on kimi-k2 divided by $0.05 on llama-3.1-8b-instant, from the price book." } ], "sources": [ { "label": "Groq pricing", "url": "https://groq.com/pricing" } ] }, { "slug": "/guardrail-cost", "url": "https://getculpa.com/guardrail-cost", "textUrl": "https://getculpa.com/guardrail-cost/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "guardrail cost", "title": "Guardrail Cost: Cheap Models Make Safety Expensive", "answer": "A guardrail call costs what it costs regardless of the model it protects, so its share of the bill rises as your main model gets cheaper. Optimising to a smaller model makes safety proportionally more expensive. Culpa, a local-first LLM cost, margin, and forecast ledger, prices every call so the check and the answer appear as separate lines.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "2.4% against 12.2%", "claim": "modelled share of a request taken by an identical safety check, on an expensive and a cheap main model", "confidence": "calculated", "source": "A modelled 500-input, 10-output check on Claude Haiku 4.5 at real rates of $1.00 and $5.00 per million costs $0.000550. A modelled protected call of 2,000 input and 500 output costs $0.0225 on Claude Opus 5 at $5.00 and $25.00, and $0.0045 on Haiku 4.5. $0.000550 / $0.0225 = 2.4% and $0.000550 / $0.0045 = 12.2%. At a million requests the check is $550.00 a month either way. Rates are published, token shapes and volume are modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/how-to-price-an-ai-saas-product", "url": "https://getculpa.com/how-to-price-an-ai-saas-product", "textUrl": "https://getculpa.com/how-to-price-an-ai-saas-product/llms.txt", "pageType": "margin", "intent": "EVALUATE", "primaryKeyword": "how to price an ai saas product", "title": "How to Price an AI SaaS Product Without Losing Money", "answer": "Price an AI SaaS product from your 90th-percentile customer cost rather than the average, then cap the tail beyond it. Usage is long-tailed, so an average-priced plan loses money on the heaviest accounts. Culpa, a local-first LLM cost, margin, and forecast ledger, tags spend to a customer and reconciles it against revenue you type in or import.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$12.00 to $260.00", "claim": "modelled monthly model cost from median to heaviest account on identical usage-based pricing", "confidence": "estimated", "source": "Both endpoints computed at real Claude Sonnet 5 introductory rates per 1k tokens from the price book, effective 2026-07-02. A range because the usage distribution is modelled." }, { "value": "$58.00", "claim": "modelled 90th-percentile monthly customer cost, the figure a plan price should clear", "confidence": "calculated", "source": "(14 x $2.00) + (3 x $10.00) per million, using real Claude Sonnet 5 introductory rates. Token volumes are modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/how-to-size-an-llm-cost-plan", "url": "https://getculpa.com/how-to-size-an-llm-cost-plan", "textUrl": "https://getculpa.com/how-to-size-an-llm-cost-plan/llms.txt", "pageType": "howto", "intent": "HOW_TO", "primaryKeyword": "size an llm cost plan", "title": "How to Size an LLM Cost Plan Before You Pick a Tier", "answer": "Size a metered plan from your token volume rather than your request count, because a credit covers one call plus one per complete 10,000 tokens. Culpa, a local-first LLM cost, margin, and forecast ledger, meters on that basis, so your tier can move when prompts grow even while traffic holds perfectly flat.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "800,000 to 1,200,000 credits", "claim": "modelled monthly credit demand for identical traffic before and after prompt growth", "confidence": "estimated", "source": "400,000 calls at 2 credits and at 3 credits, from the teardown arithmetic. A range because the token sizes and the growth are modelled." }, { "value": "39%", "claim": "modelled prompt-size increase that moved this workload a full tier at flat traffic", "confidence": "calculated", "source": "25,000 divided by 18,000 tokens per call, from the teardown arithmetic." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/idle-llm-cost", "url": "https://getculpa.com/idle-llm-cost", "textUrl": "https://getculpa.com/idle-llm-cost/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "idle llm cost", "title": "Idle LLM Cost: the Charges That Run While Nothing Does", "answer": "Some provider charges bill by elapsed time rather than per request, so they accrue while your product sits idle. Held caches, running containers and agent sessions all cost money at zero calls. Culpa, a local-first LLM cost, margin, and forecast ledger, prices calls exactly, and these belong to a different shape worth measuring separately.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$144.00 per month", "claim": "modelled cost of holding a 200,000-token Gemini context cache, making no calls", "confidence": "calculated", "source": "Gemini's pricing page, read 2026-08-05, publishes context caching storage at $1.00 per 1,000,000 tokens per hour on most models. 0.2M tokens x $1.00 x 24 hours x 30 days = $144.00. On the 2.5 Pro Priority rate of $8.10 the same holding is $1,166.40. The rates are published, the cache size and the holding period are modelled." }, { "value": "$4,147.20 per month", "claim": "modelled cost of one always-on 64 GB OpenAI container, making no calls", "confidence": "calculated", "source": "OpenAI's pricing page, read 2026-08-05, publishes container sessions at $1.92 per 20-minute session for the 64 GB tier, billed by the minute with a five-minute minimum, so $5.76 an hour, so $5.76 x 24 x 30 = $4,147.20. The rate and the tier are published, the always-on assumption is modelled." } ], "sources": [ { "label": "Gemini API pricing", "url": "https://ai.google.dev/gemini-api/docs/pricing" }, { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" }, { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/image-token-cost", "url": "https://getculpa.com/image-token-cost", "textUrl": "https://getculpa.com/image-token-cost/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "image token cost", "title": "Image Token Cost: an Image Is Input Tokens, on a Formula", "answer": "An image is billed as input tokens rather than a flat per-image fee. Claude converts it into visual tokens on a published formula, and newer models allow three times more of them for the same picture. Culpa, a local-first LLM cost, margin, and forecast ledger, prices every call from the tokens the provider reports, images included.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$1.30, $6.48 and $23.92 per thousand images", "claim": "published cost of a 1000x1000 image on two models, and a 4K image on the high-resolution tier", "confidence": "provider-reported", "source": "Anthropic's vision documentation at platform.claude.com/docs/en/build-with-claude/vision, read 2026-08-05, states each figure in its own prose. The same page publishes the formula and the token table, and re-deriving from them reproduces all three exactly: 1,296 visual tokens at $1.00 and $5.00 per million, and 4,784 tokens at $5.00 per million, each times 1,000 images. Rates match this price book's Haiku 4.5 and Opus 5 rows." }, { "value": "3.07x", "claim": "visual-token cost of one 4K image on the high-resolution tier against the standard tier", "confidence": "calculated", "source": "Anthropic's published token table gives 4,784 visual tokens for a 3840x2160 image on the high-resolution tier and 1,560 on the standard tier. 4,784 / 1,560 = 3.07, matching the same page's statement that high-resolution images can use roughly three times more visual tokens. Both token counts are published." } ], "sources": [ { "label": "Anthropic, vision and image token cost", "url": "https://platform.claude.com/docs/en/build-with-claude/vision" }, { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/input-token-creep", "url": "https://getculpa.com/input-token-creep", "textUrl": "https://getculpa.com/input-token-creep/llms.txt", "pageType": "problem", "intent": "FIX", "primaryKeyword": "input token creep", "title": "Input Token Creep: Four Percent a Month Is 60% a Year", "answer": "Input token creep is a small monthly rise in average input tokens per call across your whole workload, with no single change large enough to notice. Four percent a month compounds to about 60% in a year. Culpa, a local-first LLM cost, margin, and forecast ledger, tracks tokens per call as a series, so drift reads as a trend rather than as a surprise.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$10,000 to $16,010", "claim": "modelled monthly input spend at the start and end of a year of 4% drift, at flat call volume", "confidence": "estimated", "source": "Both endpoints at the real Claude Haiku 4.5 input rate of $0.001 per 1k from the price book, effective 2026-07-02. A range because the drift rate and volumes are modelled." }, { "value": "60%", "claim": "annual input cost increase produced by 4% monthly drift with no traffic growth", "confidence": "calculated", "source": "1.04 to the power of 12 equals 1.601, from the teardown arithmetic." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/is-my-llm-spend-normal", "url": "https://getculpa.com/is-my-llm-spend-normal", "textUrl": "https://getculpa.com/is-my-llm-spend-normal/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "is my llm spend normal", "title": "Is My LLM Spend Normal? Two Benchmarks, 191 Times Apart", "answer": "There's no industry benchmark for LLM spend, because cost tracks what the work is rather than what the company is. The two published per-unit figures differ by 191 times. Culpa, a local-first LLM cost, margin, and forecast ledger, gives you the comparison that does work, which is your own trend and your own margin.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$0.705 and $0.0037", "claim": "Anthropic's two published worked examples, an hour of agentic coding and a support conversation", "confidence": "provider-reported", "source": "Anthropic's pricing page at platform.claude.com/docs/en/docs/about-claude/pricing, read 2026-08-05, publishes both. The coding session is 50,000 input and 15,000 output on Claude Opus 5 plus one session-hour: $0.25 + $0.375 + $0.08 = $0.705, and $0.525 with caching. Support is about 3,700 tokens a conversation on Claude Haiku 4.5 at about $37.00 per 10,000, which is $0.0037 each. Re-deriving both from the published rates reproduces Anthropic's stated totals exactly. The ratio 0.705 / 0.0037 = 191." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/llama-3-1-8b-instant-pricing", "url": "https://getculpa.com/llama-3-1-8b-instant-pricing", "textUrl": "https://getculpa.com/llama-3-1-8b-instant-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "llama 3.1 8b pricing", "title": "Llama 3.1 8B Pricing on Groq: Cheap Enough to Stop Counting", "answer": "Llama 3.1 8B pricing on Groq is $0.05 per million input tokens and $0.08 per million output tokens, the cheapest rate point in this price book. A single call rounds to a hundredth of a cent, which is precisely why nobody instruments it. Culpa, a local-first LLM cost, margin, and forecast ledger, counts every call including the ones too cheap to notice.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$0.05 / $0.08 per million", "claim": "Llama 3.1 8B Instant input and output rates on Groq, the cheapest point in this price book", "confidence": "calculated", "source": "Price book row for groq/llama-3.1-8b-instant, $0.00005 and $0.00008 per 1k, effective 2026-07-02, re-verified against Groq's rendered pricing page 2026-08-01." }, { "value": "$0.000132", "claim": "modelled cost of a single 2,000-token call returning 400 tokens", "confidence": "calculated", "source": "(2 x $0.00005) + (0.4 x $0.00008) per 1k, using real Groq rates. Token counts are modelled." }, { "value": "$316.80", "claim": "modelled monthly retry waste at a 3% failure rate and a five-attempt ceiling", "confidence": "calculated", "source": "2,400,000 wasted attempts x $0.000132, from the teardown arithmetic at real Groq rates. The failure rate and call volume are modelled." } ], "sources": [ { "label": "Groq pricing", "url": "https://groq.com/pricing" } ] }, { "slug": "/llama-3-3-70b-versatile-pricing", "url": "https://getculpa.com/llama-3-3-70b-versatile-pricing", "textUrl": "https://getculpa.com/llama-3-3-70b-versatile-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "llama 3.3 70b pricing", "title": "Llama 3.3 70B Pricing on Groq: The Flattest Ratio in the Book", "answer": "Llama 3.3 70B pricing on Groq is $0.59 per million input tokens and $0.79 per million output tokens, a ratio of 1.34 and the flattest in this price book. Shortening answers, the standard cost fix, returns almost nothing at that ratio. Culpa, a local-first LLM cost, margin, and forecast ledger, prices input and output separately so the advice gets tested rather than assumed.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$0.59 / $0.79 per million", "claim": "Llama 3.3 70B Versatile input and output rates on Groq", "confidence": "calculated", "source": "Price book row for groq/llama-3.3-70b-versatile, $0.00059 and $0.00079 per 1k, effective 2026-07-02, re-verified against Groq's rendered pricing page 2026-08-01." }, { "value": "1.34x", "claim": "output rate as a multiple of input, the flattest ratio in this price book", "confidence": "calculated", "source": "$0.79 divided by $0.59, from the same price-book row." }, { "value": "14% to 36%", "claim": "modelled saving from halving output, on this model against Gemini 2.5 Flash", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at real rates for each model. A range because the token volumes are modelled." } ], "sources": [ { "label": "Groq pricing", "url": "https://groq.com/pricing" } ] }, { "slug": "/llm-ab-test-cost", "url": "https://getculpa.com/llm-ab-test-cost", "textUrl": "https://getculpa.com/llm-ab-test-cost/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "llm a/b test cost", "title": "LLM A/B Test Cost: You Shipped the Winner and Not the Bill", "answer": "An LLM A/B test compares variants on quality, and each variant also carries its own cost per call. Judging on quality alone picks a winner without pricing it. Culpa, a local-first LLM cost, margin, and forecast ledger, prices every call and tags it with the variant, so an experiment reports cost per arm beside the score.", "lastReviewed": "2026-08-03", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$28,500.00 per month", "claim": "modelled cost difference between two experiment arms at 3 million calls a month", "confidence": "calculated", "source": "Arm A at 2,000 input and 300 output tokens costs $0.0035 per call, arm B at 7,000 input and 1,200 output costs $0.0130, both on Claude Haiku 4.5 at real rates of $1.00 and $5.00 per million from the price book effective 2026-07-02. At 3,000,000 calls that's $10,500.00 against $39,000.00, a difference of $28,500.00 and a ratio of 3.71. Token shapes and call volume are modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" }, { "label": "Braintrust, interpreting eval results", "url": "https://www.braintrust.dev/docs/evaluate/interpret-results" }, { "label": "Braintrust, logging production traffic", "url": "https://www.braintrust.dev/docs/observe" } ] }, { "slug": "/llm-api-cost-calculator", "url": "https://getculpa.com/llm-api-cost-calculator", "textUrl": "https://getculpa.com/llm-api-cost-calculator/llms.txt", "pageType": "calculator", "intent": "HOW_TO", "primaryKeyword": "llm api cost calculator", "title": "LLM API Cost Calculator: Exact Decimal, Not Floating Point", "answer": "This LLM API cost calculator prices your own monthly token volumes against any model in Culpa's price book, splitting uncached input, cached input and output because they bill at three different rates. Culpa, a local-first LLM cost, margin, and forecast ledger, computes every figure here in exact decimal rather than floating point, so a total matches an invoice to the cent.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": "single", "costFigures": [ { "value": "$610.00", "claim": "modelled monthly cost of the teardown workload priced across all three rates", "confidence": "calculated", "source": "(200 x $2.00) + (50 x $0.20) + (20 x $10.00) per million, using real Claude Sonnet 5 rates from the price book, effective 2026-07-02. Token volumes are modelled." }, { "value": "$610.00 to $1,010.00", "claim": "modelled monthly cost of the same 270M tokens under two different splits", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at real rates. A range because the split between input and output is modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/llm-cost-accrual", "url": "https://getculpa.com/llm-cost-accrual", "textUrl": "https://getculpa.com/llm-cost-accrual/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "llm cost accrual", "title": "LLM Cost Accrual: Closing the Month Before the Invoice Arrives", "answer": "LLM cost accrual is booking model spend in the month it was incurred, before the provider invoice arrives. Accruing from last month's total under-books any service still growing. Culpa, a local-first LLM cost, margin, and forecast ledger, prices every call as it happens, so the month's figure exists on the last day of the month rather than a week later.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$5,000, or 28.6%", "claim": "modelled under-accrual from booking a growing month at last month's actual", "confidence": "calculated", "source": "March at 8,000M input and 900M output on Claude Haiku 4.5 at real rates of $1.00 and $5.00 per million from the price book effective 2026-07-02 is $8,000 + $4,500 = $12,500. April at 11,000M and 1,300M is $11,000 + $6,500 = $17,500. Accruing April at March's actual leaves $5,000 unbooked, which is 28.6% of $17,500. Token volumes are modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/llm-cost-anomaly-detection", "url": "https://getculpa.com/llm-cost-anomaly-detection", "textUrl": "https://getculpa.com/llm-cost-anomaly-detection/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "llm cost anomaly detection", "title": "LLM Cost Anomaly Detection: The Aggregate Has a Noise Floor", "answer": "LLM cost anomaly detection flags spend that departs from its own pattern. Run it on a company total and it inherits a noise floor, because a single customer tripling can be a fraction of a percent of the whole. Culpa, a local-first LLM cost, margin, and forecast ledger, prices every call, so a detector can watch a customer rather than a company.", "lastReviewed": "2026-08-03", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "0.2%", "claim": "a modelled customer tripling their spend, expressed as a share of the company total", "confidence": "calculated", "source": "A $100.00 movement against a modelled $50,000.00 monthly total = 0.2%, against 200% when measured against that customer's own $50.00 baseline. Spend figures are modelled and priced on Claude Haiku 4.5 at real rates of $1.00 and $5.00 per million from the price book effective 2026-07-02. The pair is the finding rather than either number alone." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/llm-cost-by-user-count", "url": "https://getculpa.com/llm-cost-by-user-count", "textUrl": "https://getculpa.com/llm-cost-by-user-count/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "llm cost for 100 users", "title": "LLM Cost for 100 Users: Price One User, Then Multiply", "answer": "Cost for a user count is one bundle priced once and multiplied, so the useful number is cost per user rather than the total. The total moves for two different reasons and won't tell you which. Culpa, a local-first LLM cost, margin, and forecast ledger, prices every call and attributes it to the user who made it.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": "scale", "costFigures": [ { "value": "$1.00 per user per month", "claim": "modelled cost of one user at 500,000 input and 100,000 output tokens on Claude Haiku 4.5", "confidence": "calculated", "source": "500,000 input at $1.00 per million is $0.50 and 100,000 output at $5.00 per million is $0.50, giving $1.00, using the price book's real Claude Haiku 4.5 rates effective 2026-07-02. At 100 users that comes to $100.00 and at 1,000 users $1,000.00. The same user on Claude Sonnet 5 at $2.00 and $10.00 costs $2.00. Rates are published, the per-user token volumes and user counts are modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/llm-cost-chargeback", "url": "https://getculpa.com/llm-cost-chargeback", "textUrl": "https://getculpa.com/llm-cost-chargeback/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "llm cost chargeback", "title": "LLM Cost Chargeback: Splitting One API Key Between Teams", "answer": "LLM cost chargeback means allocating shared model spend to the internal teams that caused it, using recorded per-call attribution rather than a headcount split. One key gives you one number, and a number resists being divided fairly. Culpa, a local-first LLM cost, margin, and forecast ledger, tags every call with its team and prices it individually.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$11,400 per month", "claim": "modelled amount a headcount split undercharges the heaviest team on a $24,000 month", "confidence": "calculated", "source": "19,000M input and 1,000M output tokens on Claude Haiku 4.5 at real rates of $1.00 and $5.00 per million from the price book effective 2026-07-02 gives $19,000 + $5,000 = $24,000. Split across 30 people that's $800 a head, so a team of 6 is charged $4,800 against modelled measured usage of $16,200, a shortfall of $11,400. Token volumes, team sizes and the usage split are modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/llm-cost-forecast-calculator", "url": "https://getculpa.com/llm-cost-forecast-calculator", "textUrl": "https://getculpa.com/llm-cost-forecast-calculator/llms.txt", "pageType": "problem", "intent": "IMPROVE", "primaryKeyword": "llm cost forecast", "title": "LLM Cost Forecast Calculator: A Range, Not a Number", "answer": "An LLM cost forecast projects next month's spend from your own history, and it has to be a range because a single number hides how much your months disagree. This calculator prices three months you enter and projects the next one. Culpa, a local-first LLM cost, margin, and forecast ledger, forecasts from your full ledger and scores every forecast afterwards.", "lastReviewed": "2026-08-02", "effectiveDate": "2026-07-02", "calculatorMode": "forecast", "costFigures": [ { "value": "$294.74 to $532.00", "claim": "forecast band for a modelled product whose growth lurched from +90.0% to +5.3%", "confidence": "estimated", "source": "Computed by this page's calculator from months of 100M, 190M and 200M tokens at a 10% output share, priced on Claude Haiku 4.5 at real rates of $1.00 and $5.00 per million from the price book effective 2026-07-02. Observed growth steps are +90.0% then +5.3%, so the bounds are last month's 200M grown by each. Token volumes are modelled. Published as a range because it's an estimate, and because two growth steps support a range and nothing narrower." }, { "value": "$307.69 to $308.85", "claim": "forecast band for a modelled product at the same last-month volume, growing steadily", "confidence": "estimated", "source": "Same method and same rates, from months of 165M, 182M and 200M tokens at a 10% output share, giving growth steps of +10.3% then +9.9%. Both products ended last month at 200M tokens and both priced at $280.00. The steady band spans $1.16 and the lurching one spans $237.26, a factor of 204, which is the point of the pair." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/llm-cost-in-datadog", "url": "https://getculpa.com/llm-cost-in-datadog", "textUrl": "https://getculpa.com/llm-cost-in-datadog/llms.txt", "pageType": "problem", "intent": "HOW_TO", "primaryKeyword": "llm cost in datadog", "title": "LLM Cost in Datadog: Two Products, Two Different Answers", "answer": "Datadog answers LLM cost twice. LLM Observability monitors token usage and cost across agents in real time, and Cloud Cost Management ingests OpenAI as a SaaS spend source and forecasts it. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each call and sets it against the revenue the customer who triggered it pays you.", "lastReviewed": "2026-08-03", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$200.00 per month", "claim": "Datadog Cloud Cost Management Pro on a modelled $40,000 of monthly cloud and SaaS spend", "confidence": "calculated", "source": "$5 per $1,000 in cloud and SaaS spend per month, published on datadoghq.com/pricing and read 2026-08-03, applied to a modelled $40,000 monthly spend: 40 x $5.00 = $200.00, which is 0.5% of the spend covered. The rate is published and the spend figure is modelled." } ], "sources": [ { "label": "Datadog LLM Observability", "url": "https://www.datadoghq.com/product/llm-observability/" }, { "label": "Datadog pricing", "url": "https://www.datadoghq.com/pricing/" } ] }, { "slug": "/llm-cost-in-dynatrace", "url": "https://getculpa.com/llm-cost-in-dynatrace", "textUrl": "https://getculpa.com/llm-cost-in-dynatrace/llms.txt", "pageType": "problem", "intent": "HOW_TO", "primaryKeyword": "llm cost in dynatrace", "title": "LLM Cost in Dynatrace: A Predicted Rise Beats No Warning, and Beats No Forecast", "answer": "Dynatrace AI Observability assesses token usage and cost across models and says it can predict cost increases so you can act before they land. Culpa, a local-first LLM cost, margin, and forecast ledger, projects next month from your own history as a range and sets each customer's spend against the revenue they pay, which is a different question from spotting a rise.", "lastReviewed": "2026-08-03", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$312.00 to $426.24", "claim": "modelled monthly range for the same workload, flat against continued growth", "confidence": "estimated", "source": "Both bounds run from the same two observed weeks, $60.00 then $84.00, priced on Claude Haiku 4.5 at real rates of $1.00 and $5.00 per million from the price book effective 2026-07-02. If growth stops after week 2 the month is $60.00 + $84.00 + $84.00 + $84.00 = $312.00. If 40% weekly growth continues the month reaches $60.00 + $84.00 + $117.60 + $164.64 = $426.24. Every volume is modelled and the figure is published as a range because it's an estimate." } ], "sources": [ { "label": "Dynatrace AI Observability", "url": "https://www.dynatrace.com/solutions/ai-observability/" }, { "label": "Dynatrace AI observability docs", "url": "https://docs.dynatrace.com/docs/observe/dynatrace-for-ai-observability" } ] }, { "slug": "/llm-cost-in-grafana", "url": "https://getculpa.com/llm-cost-in-grafana", "textUrl": "https://getculpa.com/llm-cost-in-grafana/llms.txt", "pageType": "problem", "intent": "HOW_TO", "primaryKeyword": "llm cost in grafana", "title": "LLM Cost in Grafana: Per-Agent Dollars, and a Token Bill of Its Own", "answer": "Grafana Cloud's AI observability does real-time spend tracking and per-agent cost breakdown in USD, which is more cost attribution than several LLM-native tools publish. Culpa, a local-first LLM cost, margin, and forecast ledger, prices calls from a versioned price book and joins them to the revenue each customer pays, which is the column Grafana has no reason to hold.", "lastReviewed": "2026-08-03", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "375%", "claim": "modelled rise in ingested telemetry from adding full LLM tracing, with model spend unchanged", "confidence": "calculated", "source": "20 GB ingested before tracing against 95 GB after, both modelled, giving a 375% increase, while modelled model spend of $80.00 a month on Claude Haiku 4.5 at real rates of $1.00 and $5.00 per million from the price book effective 2026-07-02 stays flat. Every volume here is modelled. The point is that the two meters move independently, not the size of either." } ], "sources": [ { "label": "Grafana Cloud AI observability", "url": "https://grafana.com/docs/grafana-cloud/monitor-applications/ai-observability/" }, { "label": "Grafana pricing", "url": "https://grafana.com/pricing/" } ] }, { "slug": "/llm-cost-in-new-relic", "url": "https://getculpa.com/llm-cost-in-new-relic", "textUrl": "https://getculpa.com/llm-cost-in-new-relic/llms.txt", "pageType": "problem", "intent": "HOW_TO", "primaryKeyword": "llm cost in new relic", "title": "LLM Cost in New Relic: Your Traces Bill by the Gigabyte", "answer": "New Relic AI Monitoring gives visibility into AI performance and token usage, and New Relic bills data ingest per gigabyte, so instrumenting LLM calls raises a second bill. Culpa, a local-first LLM cost, margin, and forecast ledger, prices calls from a versioned price book on your own infrastructure and sets them against the revenue each customer pays.", "lastReviewed": "2026-08-03", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$120.00 per month", "claim": "modelled New Relic ingest for 400 GB of LLM trace text on standard data pricing", "confidence": "calculated", "source": "400 GB modelled ingest minus the published 100 GB free allowance = 300 GB, at the published $0.40 per GB standard rate = $120.00, or $180.00 at the $0.60 Data Plus rate. Allowance and rates from newrelic.com/pricing read 2026-08-03. Trace volume and payload size are modelled." }, { "value": "$0.40 to $0.60 per GB", "claim": "New Relic's published data ingest rates beyond the free allowance", "confidence": "provider-reported", "source": "newrelic.com/pricing, read 2026-08-03. 100 GB free each month, then $0.40 per GB on standard data ingest and $0.60 per GB on Data Plus, with an additional $0.05 per GB per month to store data in the European Union." } ], "sources": [ { "label": "New Relic AI Monitoring", "url": "https://newrelic.com/platform/ai-monitoring" }, { "label": "New Relic pricing", "url": "https://newrelic.com/pricing" } ] }, { "slug": "/llm-cost-per-deploy", "url": "https://getculpa.com/llm-cost-per-deploy", "textUrl": "https://getculpa.com/llm-cost-per-deploy/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "llm cost per deploy", "title": "LLM Cost per Deploy: Unit Cost Moves and the Total Hides It", "answer": "Tracking LLM cost per deploy means recording which release produced each call, then comparing cost per call across releases rather than comparing monthly totals. Traffic and code change at the same time, so a total can't separate them. Culpa, a local-first LLM cost, margin, and forecast ledger, tags every call with its release and prices it individually.", "lastReviewed": "2026-08-03", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$280.00", "claim": "modelled monthly movement from a 15% unit-cost regression, masked by a 10% traffic fall", "confidence": "calculated", "source": "Per call, 2,000 input and 400 output tokens cost $0.0040, and 2,600 input and 400 output cost $0.0046, both on Claude Haiku 4.5 at real rates of $1.00 and $5.00 per million from the price book effective 2026-07-02, a 15.0% rise. At 2,000,000 calls before and 1,800,000 after, $8,000.00 against $8,280.00, a $280.00 or 3.5% movement. Token shapes and call volumes are modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/llm-cost-spike-after-launch", "url": "https://getculpa.com/llm-cost-spike-after-launch", "textUrl": "https://getculpa.com/llm-cost-spike-after-launch/llms.txt", "pageType": "problem", "intent": "AUDIT", "primaryKeyword": "llm cost spike after launch", "title": "LLM Cost Spike After Launch: What Actually Drives It", "answer": "A launch spike is rarely proportional to traffic. New users explore more per session than retained ones, they arrive on your cheapest tier, and any unbounded loop meets its worst input at once. Culpa, a local-first LLM cost, margin, and forecast ledger, forecasts a launch from your own usage curve and prices it before you ship.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$440 to $1,730", "claim": "modelled launch-month spend, naive forecast versus cohort-based forecast", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at real GPT-5.4 rates from the price book, effective 2026-07-02. A range because the first-week usage multiple is modelled, not measured." }, { "value": "$0.22", "claim": "modelled steady-state model cost per active user per month", "confidence": "calculated", "source": "(40 x $0.0025) + (8 x $0.015) using GPT-5.4 rates per 1k tokens from the price book. Per-user token volumes are modelled." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/llm-eval-job-cost", "url": "https://getculpa.com/llm-eval-job-cost", "textUrl": "https://getculpa.com/llm-eval-job-cost/llms.txt", "pageType": "problem", "intent": "FIX", "primaryKeyword": "llm eval job cost", "title": "LLM Eval Job Cost: The Line Item Nobody Watches", "answer": "Evaluation runs bill at production rates while serving no customer, and they scale with test cases times prompt variants times a judge call. That product grows fast. Culpa, a local-first LLM cost, margin, and forecast ledger, tags eval traffic as its own feature, so it stops hiding inside your unattributed spend.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$14.55", "claim": "modelled cost of one full eval run across 200 cases and 3 variants with a model judge", "confidence": "calculated", "source": "600 test calls at $0.00975 plus 600 judge calls at $0.0145, using real GPT-5.4 and GPT-5.5 rates per 1k tokens from the price book, effective 2026-07-02. Case counts and token sizes are modelled." }, { "value": "60%", "claim": "share of the modelled eval run cost taken by the judge rather than the model under test", "confidence": "calculated", "source": "$8.70 judge subtotal divided by the $14.55 run total, from the teardown arithmetic." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/llm-fast-mode-pricing", "url": "https://getculpa.com/llm-fast-mode-pricing", "textUrl": "https://getculpa.com/llm-fast-mode-pricing/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "llm fast mode pricing", "title": "LLM Fast Mode Pricing: 2x, Except Where It Runs 2.5x", "answer": "Fast mode is a premium latency tier that bills the same tokens at a higher rate. OpenAI charges double on most models and 2.5 times on GPT-5.5, and Anthropic charges double on two Opus models. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each call from a versioned price book, so a tier change shows up as a cost rather than a setting.", "lastReviewed": "2026-08-03", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$2,000.00 per month", "claim": "modelled GPT-5.5 volume on fast mode, against $800.00 standard and $400.00 batched", "confidence": "calculated", "source": "100M input and 10M output tokens at OpenAI's published GPT-5.5 rates read 2026-08-03: fast mode $12.50 and $75.00 per million gives $2,000.00, standard $5.00 and $30.00 gives $800.00, batch at half of standard gives $400.00. Token volume is modelled and all three rate pairs are published." }, { "value": "2.5x", "claim": "OpenAI's published fast-mode multiplier on GPT-5.5, against 2x on every other model listed", "confidence": "provider-reported", "source": "developers.openai.com/api/docs/pricing, read 2026-08-03. GPT-5.5 standard is $5.00 and $30.00 per million and its fast-mode row is $12.50 and $75.00, which is 2.5x on both. The GPT-5.6 family and GPT-5.4 are 2x on both. Reported because a single multiplier applied across a family would be wrong on the model many teams are still running." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" }, { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/llm-model-mix-drift", "url": "https://getculpa.com/llm-model-mix-drift", "textUrl": "https://getculpa.com/llm-model-mix-drift/llms.txt", "pageType": "problem", "intent": "FIX", "primaryKeyword": "llm model mix drift", "title": "LLM Model Mix Drift: The Bill Rises With No Code Change", "answer": "Model mix drift is traffic shifting toward pricier models without anyone changing the code, through fallbacks, routing rules or a default that moved. Volume holds and the bill climbs. Culpa, a local-first LLM cost, margin, and forecast ledger, records the model that served every call, so a mix shift separates cleanly from a price change.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "30.6%", "claim": "modelled bill increase from a tenth of traffic moving to a pricier model at identical volume", "confidence": "calculated", "source": "($10.45 - $8.00) / $8.00, using real Gemini 2.5 Flash and Gemini 2.5 Pro rates per 1k tokens from the price book, effective 2026-07-02. Token volumes are modelled." }, { "value": "$8.00 to $10.45", "claim": "modelled monthly spend before and after the mix shift, same token volume", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at real Gemini rates. A range because the traffic split is modelled." } ], "sources": [ { "label": "Gemini API pricing", "url": "https://ai.google.dev/gemini-api/docs/pricing" } ] }, { "slug": "/llm-observability-consolidation", "url": "https://getculpa.com/llm-observability-consolidation", "textUrl": "https://getculpa.com/llm-observability-consolidation/llms.txt", "pageType": "problem", "intent": "EVALUATE", "primaryKeyword": "llm observability consolidation", "title": "LLM Observability Consolidation: Six Tools That Changed Hands", "answer": "Six named LLM observability tools are no longer independent, from Langfuse and Helicone through to one shut down outright. This consolidation moves your cost history along with them. Culpa, a local-first LLM cost, margin, and forecast ledger, prices and forecasts that history in a database on your own infrastructure, so an owner change never sets how many months you keep.", "lastReviewed": "2026-08-03", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$170.00 per month", "claim": "the published price difference between 90 days and 3 years of retained history on one vendor", "confidence": "calculated", "source": "langfuse.com/pricing, read 2026-08-03. Pro at $199.00 minus Core at $29.00 = $170.00 a month, and the feature table on that page gives Core 90 days of data access against Pro's 3 years. Both prices are Langfuse's own published list prices. The subtraction is the only arithmetic here and it's shown in full." } ], "sources": [ { "label": "ClickHouse acquires Langfuse", "url": "https://clickhouse.com/blog/clickhouse-acquires-langfuse-open-source-llm-observability" }, { "label": "Langfuse joins ClickHouse", "url": "https://langfuse.com/blog/joining-clickhouse" }, { "label": "Langfuse pricing", "url": "https://langfuse.com/pricing" }, { "label": "Helicone: joining Mintlify", "url": "https://www.helicone.ai/blog/joining-mintlify" }, { "label": "Traceloop joining ServiceNow", "url": "https://www.traceloop.com/blog/traceloop-is-joining-servicenow" }, { "label": "Humanloop", "url": "https://humanloop.com" }, { "label": "Migrating from Humanloop", "url": "https://humanloop.com/docs/v5/guides/migrating-from-humanloop" }, { "label": "keywordsai.co", "url": "https://www.keywordsai.co" }, { "label": "Respan", "url": "https://www.respan.ai" }, { "label": "Literal AI migration guide", "url": "https://docs.literalai.com/more/migration-guide" } ] }, { "slug": "/llm-pricing-modifiers", "url": "https://getculpa.com/llm-pricing-modifiers", "textUrl": "https://getculpa.com/llm-pricing-modifiers/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "llm pricing modifiers", "title": "LLM Pricing Modifiers: Which Ones Stack and Which Ones Conflict", "answer": "A published per-token rate is a starting point that four modifiers move: prompt caching, batch processing, fast mode and data residency. Most multiply together and one pair is mutually exclusive. Culpa, a local-first LLM cost, margin, and forecast ledger, prices every call at the rate it was actually billed rather than at the headline one.", "lastReviewed": "2026-08-03", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$0.50 to $11.00", "claim": "the range one million Opus 4.8 input tokens can cost, depending on which modifiers are set", "confidence": "calculated", "source": "1M input tokens at Anthropic's published Opus $5.00 per million from the price book effective 2026-07-02, multiplied by published modifiers read from platform.claude.com on 2026-08-03: cached read 0.10x gives $0.50, batch 0.50x gives $2.50, fast mode 2x with regional residency 1.10x gives $11.00. Every multiplier is published. The combinations shown are documented as legal, and fast mode with batch is documented as unavailable." }, { "value": "10%", "claim": "the regional and multi-region endpoint premium, published by both OpenAI and Anthropic", "confidence": "provider-reported", "source": "Anthropic: regional and multi-region endpoints include a 10% premium over global endpoints, applying from Claude Sonnet 4.5, Haiku 4.5 and Opus 4.5 onward. OpenAI: regional processing endpoints are charged a 10% uplift for eligible models released on or after 2026-03-05. Both read on 2026-08-03." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" }, { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/llm-retry-loop-cost", "url": "https://getculpa.com/llm-retry-loop-cost", "textUrl": "https://getculpa.com/llm-retry-loop-cost/llms.txt", "pageType": "problem", "intent": "FIX", "primaryKeyword": "llm retry loop cost", "title": "LLM Retry Loop Cost: Why Failed Calls Still Bill You", "answer": "A retry loop costs you the full rate on every attempt, because a provider bills tokens it generated even when your code discarded the response. Culpa, a local-first LLM cost, margin, and forecast ledger, meters failed and retried calls alongside successful ones, so your retry waste shows up as a number instead of an unexplained gap.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$0.042", "claim": "billed cost of one usable result after two rejected attempts", "confidence": "calculated", "source": "Three attempts at (3 x $0.002) + (0.8 x $0.010), using Claude Sonnet 5 introductory rates per 1k tokens from the price book, effective 2026-07-02." }, { "value": "3x", "claim": "multiple of intended cost paid for a single usable result at a three-attempt retry", "confidence": "calculated", "source": "$0.042 billed divided by the $0.014 single-attempt cost, from the teardown arithmetic." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/llm-streaming-cost-tracking", "url": "https://getculpa.com/llm-streaming-cost-tracking", "textUrl": "https://getculpa.com/llm-streaming-cost-tracking/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "llm streaming cost tracking", "title": "Streaming LLM Cost Tracking: Usage Lives at the End, Not the Start", "answer": "Streaming responses report usage across two events rather than one body, and the numbers can differ within a single request. Reading the opening event alone under-counts, and summing the closing ones double-counts. Culpa, a local-first LLM cost, margin, and forecast ledger, records the final figures and prices each call from them.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-08-01", "calculatorMode": null, "costFigures": [ { "value": "8,003 input tokens", "claim": "difference between the two input counts in Anthropic's own published streaming example", "confidence": "provider-reported", "source": "Anthropic's streaming documentation at platform.claude.com/docs/en/build-with-claude/streaming, read 2026-08-05. Its worked web-search example on claude-opus-5 publishes input_tokens of 2,679 in the message_start event and 10,682 in the final message_delta event of the same request. Both figures are Anthropic's, quoted as published. The difference of 8,003 is 74.9% of 10,682." }, { "value": "$0.013470 against $0.066160", "claim": "cost of that one published call, read from the first usage event and from the last", "confidence": "calculated", "source": "Applying the price book's real claude-opus-5 rates of $5.00 and $25.00 per million, effective 2026-08-01, to Anthropic's own published token counts. The message_start event reports 2,679 input and 3 output, giving $0.013395 + $0.000075 = $0.013470. The final message_delta reports 10,682 and 510, giving $0.053410 + $0.012750 = $0.066160, a factor of 4.91. Nothing here is modelled. Both token counts and both rates are published. Token cost only: the same published example also performed one web search, billed separately at $10 per 1,000, which takes the true total to $0.076160." } ], "sources": [ { "label": "Anthropic, streaming messages", "url": "https://platform.claude.com/docs/en/build-with-claude/streaming" }, { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/llm-tokenizer-change-cost", "url": "https://getculpa.com/llm-tokenizer-change-cost", "textUrl": "https://getculpa.com/llm-tokenizer-change-cost/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "llm tokenizer change cost", "title": "An LLM Tokenizer Change Is a Price Rise With No Price Change", "answer": "A tokenizer change alters how much text becomes one billable token, so identical prompts cost different amounts at an identical published rate. Anthropic states Claude 4.7 and later produce about 30% more tokens for the same text. Culpa, a local-first LLM cost, margin, and forecast ledger, meters the tokens each call actually reported rather than the rate its model advertises.", "lastReviewed": "2026-08-03", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$450.00 per month", "claim": "modelled increase from a tokenizer change alone, with the published rate held constant", "confidence": "calculated", "source": "200M input and 20M output tokens at $5.00 and $25.00 per million, the published Opus rate from the price book effective 2026-07-02, against the same text at 30% more tokens: $1,500.00 becomes $1,950.00. The 30% figure is Anthropic's own, from platform.claude.com pricing read 2026-08-03. Token volumes are modelled and the 30% is a stated approximation, so treat the shape rather than the figure as transferable." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/long-context-cost", "url": "https://getculpa.com/long-context-cost", "textUrl": "https://getculpa.com/long-context-cost/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "long context cost", "title": "Long Context Cost: No Premium, and 100 Times the Bill", "answer": "Anthropic charges no premium for long context: a 900,000-token request bills at the same per-token rate as a 9,000-token one. The rate holds and the token count doesn't, so the same request costs a hundred times more. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each call from its own token count.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$1.80 against $0.0180", "claim": "input cost of a 900k-token request and a 9k-token request on Claude Sonnet 5, at the same rate", "confidence": "calculated", "source": "Anthropic's pricing page, read 2026-08-05, states that Claude 4.6 and later include the full 1M-token context window at standard pricing and that a 900k-token request is billed at the same per-token rate as a 9k-token request. At Claude Sonnet 5's introductory $2.00 per million from the price book effective 2026-07-02, 900,000 tokens is $1.80 and 9,000 is $0.0180, a factor of exactly 100. The same 900,000 tokens on Claude Opus 5 at $5.00 per million is $4.50. Rates and the no-premium statement are published, the request sizes come from Anthropic's own example." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/model-mix-cost-teardown", "url": "https://getculpa.com/model-mix-cost-teardown", "textUrl": "https://getculpa.com/model-mix-cost-teardown/llms.txt", "pageType": "teardown", "intent": "EXAMPLES", "primaryKeyword": "model mix cost teardown", "title": "Model Mix Cost Teardown: The Cheap Model Did More Work", "answer": "This is a real month of Anthropic API usage, measured rather than modelled. The cheaper model processed 20% more input tokens and 12 times more output than the frontier model beside it, for 55% of the cost. Culpa, a local-first LLM cost, margin, and forecast ledger, attributes spend per model, which is what turns a model mix into a decision rather than a habit.", "lastReviewed": "2026-08-03", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "54.7%", "claim": "share of one real month's model cost taken by the frontier model on this account", "confidence": "provider-reported", "source": "Anthropic usage export for the month, per-model cost aggregated by scripts/analyse-usage-export.py and expressed as a share of the month. Account identifiers stripped on load." }, { "value": "30.3%", "claim": "share of the same month taken by the small model, which processed more tokens", "confidence": "provider-reported", "source": "The same export and aggregation, per-model cost as a share of the month." }, { "value": "12.2x", "claim": "measured ratio of the small model's output tokens to the frontier model's", "confidence": "calculated", "source": "Haiku 4.5 output tokens divided by Opus 4.8 output tokens, from the provider's usage export. Published as a ratio, because the counts would recover a withheld total against a public rate card." }, { "value": "3.4% and 34.6%", "claim": "output as a share of input for the frontier model and the small model", "confidence": "calculated", "source": "Output tokens divided by input tokens for each model, from the export's own counts, published as shares for the same reason the counts aren't." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/model-switch-cost-calculator", "url": "https://getculpa.com/model-switch-cost-calculator", "textUrl": "https://getculpa.com/model-switch-cost-calculator/llms.txt", "pageType": "calculator", "intent": "HOW_TO", "primaryKeyword": "model switch cost calculator", "title": "Model Switch Cost Calculator: Price the Move Before You Make It", "answer": "This model switch cost calculator prices two models against the same token volumes and reports which one your own mix favours, plus the output-to-input ratio where the two tie. Culpa, a local-first LLM cost, margin, and forecast ledger, computes both totals in exact decimal, so a migration gets decided on arithmetic rather than on a headline rate.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": "compare", "costFigures": [ { "value": "0.85", "claim": "output-to-input ratio at which the two teardown models cost the same", "confidence": "calculated", "source": "($0.59 - $0.20) divided by ($1.25 - $0.79), using real GPT-5.4 nano and Groq Llama 3.3 70B Versatile rates per million from the price book, effective 2026-07-02." }, { "value": "$61.52 to $133.80", "claim": "modelled monthly cost across two features and two models, cheapest to dearest", "confidence": "estimated", "source": "The four totals in the teardown arithmetic at real rates. A range because the token volumes are modelled." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" }, { "label": "Groq pricing", "url": "https://groq.com/pricing" } ] }, { "slug": "/model-upgrade-cost-increase", "url": "https://getculpa.com/model-upgrade-cost-increase", "textUrl": "https://getculpa.com/model-upgrade-cost-increase/llms.txt", "pageType": "problem", "intent": "FIX", "primaryKeyword": "model upgrade cost increase", "title": "Model Upgrade Cost Increase: Same Workload, New Rate", "answer": "A model upgrade cost increase is the same workload priced at a new rate, and it happens whether or not you changed anything. Claude Sonnet 5's introductory pricing ends on 2026-09-01 and rises 50%. Culpa, a local-first LLM cost, margin, and forecast ledger, prices every call from a dated price book, so a rate change lands as a forecast rather than as an invoice.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$3,200 to $4,800", "claim": "modelled monthly cost of one unchanged workload on Claude Sonnet 5, before and after 2026-09-01", "confidence": "estimated", "source": "Both endpoints use real Claude Sonnet 5 rows from the price book, $2.00 and $10.00 per million effective 2026-07-02 and $3.00 and $15.00 effective 2026-09-01. A range because the token volumes are modelled." }, { "value": "$1,600", "claim": "modelled monthly increase from the scheduled Sonnet 5 rate change, at unchanged volume", "confidence": "calculated", "source": "$4,800 minus $3,200, from the teardown arithmetic at real price book rates." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/moonshotai-kimi-k2-instruct-0905-pricing", "url": "https://getculpa.com/moonshotai-kimi-k2-instruct-0905-pricing", "textUrl": "https://getculpa.com/moonshotai-kimi-k2-instruct-0905-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "kimi k2 pricing", "title": "Kimi K2 Pricing on Groq: The Cached Rate Nobody Switches On", "answer": "Kimi K2 pricing on Groq is $1.00 per million input tokens and $3.00 per million output tokens, with a cached read at exactly half the input rate. On agentic traffic, where tool definitions repeat on every step, that discount is worth about a quarter of the bill. Culpa, a local-first LLM cost, margin, and forecast ledger, prices cached and uncached input separately so the unclaimed part is visible.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$1.00 / $3.00 per million", "claim": "Kimi K2 Instruct input and output rates on Groq, with a cached read at half input", "confidence": "calculated", "source": "Price book row for groq/moonshotai/kimi-k2-instruct-0905, $0.001 and $0.003 per 1k with cached $0.0005, effective 2026-07-02, re-verified against Groq's rendered pricing page 2026-08-01." }, { "value": "25%", "claim": "modelled saving from caching a repeated agent prefix making up 76% of input", "confidence": "calculated", "source": "$19.00 saved against a $77.00 uncached total, from the teardown arithmetic at real Groq rates. The repeated share is modelled." } ], "sources": [ { "label": "Groq pricing", "url": "https://groq.com/pricing" } ] }, { "slug": "/multi-agent-cost-calculator", "url": "https://getculpa.com/multi-agent-cost-calculator", "textUrl": "https://getculpa.com/multi-agent-cost-calculator/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "multi agent system cost calculator", "title": "Multi-Agent Cost Calculator: Price a Step, Multiply by Depth", "answer": "A multi-agent system run costs its per-step bundle times the number of steps it takes, and that count is decided by the model at runtime rather than by your configuration. Culpa, a local-first LLM cost, margin, and forecast ledger, prices every call and groups them by run, so a run reports its own total and its own depth.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": "scale", "costFigures": [ { "value": "$0.0240 a step", "claim": "modelled cost of one agent step at 8,000 input and 800 output tokens on Claude Sonnet 5", "confidence": "calculated", "source": "8,000 input at $2.00 per million is $0.0160 and 800 output at $10.00 per million is $0.0080, giving $0.0240, using the price book's real Claude Sonnet 5 introductory rates effective 2026-07-02. Multiplied by modelled depths that gives $0.0720 at 3 steps, $0.1680 at 7 and $0.4800 at 20, a ratio of 6.67 between the deepest and shallowest. Rates are published, the step shape and depths are modelled, and the figures exclude context growth across steps." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/multi-agent-cost-tracking", "url": "https://getculpa.com/multi-agent-cost-tracking", "textUrl": "https://getculpa.com/multi-agent-cost-tracking/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "multi agent cost tracking", "title": "Multi-Agent Cost Tracking: You Pay for the Context Every Hop", "answer": "Multi-agent cost tracking means attributing spend across agents that delegate to each other, including the context each hand-off re-sends. Delegation multiplies input tokens rather than dividing work. Culpa, a local-first LLM cost, margin, and forecast ledger, prices every call and rolls it up to the agent, the run and the customer that caused it.", "lastReviewed": "2026-08-03", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$0.0500", "claim": "modelled cost of one four-agent task, of which $0.0320 is the same context sent four times", "confidence": "calculated", "source": "Arithmetic shown in full on this page. 8,000 shared context tokens x 4 agents plus 1,500 own instructions x 4 = 38,000 input tokens, and 600 output x 4 = 2,400 output tokens, priced on Claude Haiku 4.5 at real rates of $1.00 and $5.00 per million from the price book effective 2026-07-02. Every token count and the hop count are modelled." }, { "value": "64%", "claim": "share of that modelled task spent re-sending context across hops", "confidence": "calculated", "source": "$0.0320 of a $0.0500 task total = 64.0%, from the same modelled shape and the same real rates. Reported as a share because the share is the transferable finding: the architecture rather than the workload sets most of the bill." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/multi-tenant-llm-cost", "url": "https://getculpa.com/multi-tenant-llm-cost", "textUrl": "https://getculpa.com/multi-tenant-llm-cost/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "multi tenant llm cost", "title": "Multi-Tenant LLM Cost: Your Cache Doesn't Scale the Way You Think", "answer": "A prompt cache is keyed on content, so a system prompt every tenant shares still gets cached separately wherever the surrounding context differs. Cache cost then scales with tenant count rather than staying fixed. Culpa, a local-first LLM cost, margin, and forecast ledger, prices every call and attributes it to the tenant that made it.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$3,600.00 against $36.00 per month", "claim": "modelled cache-write cost for one shared system prompt across 100 tenants against one", "confidence": "calculated", "source": "A modelled 20,000-token system prompt on Claude Sonnet 5, where the 5-minute cache write rate is $2.50 per million from the price book effective 2026-07-02, being 1.25x the $2.00 base input rate Anthropic publishes. One write costs $0.0500, so hourly refreshes across 30 days cost $36.00 for one cache entry and $3,600.00 for 100. Rates and the cache multiplier are published, the prompt size, tenant count and refresh rate are modelled." } ], "sources": [ { "label": "Anthropic, prompt caching", "url": "https://platform.claude.com/docs/en/build-with-claude/prompt-caching" }, { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/openai-api-cost", "url": "https://getculpa.com/openai-api-cost", "textUrl": "https://getculpa.com/openai-api-cost/llms.txt", "pageType": "provider_cost", "intent": "EVALUATE", "primaryKeyword": "openai api cost", "title": "OpenAI API Cost: A 150x Spread From Nano to Pro", "answer": "OpenAI API cost runs from $0.20 per million input tokens on nano to $30.00 on the Pro tier, a 150-fold spread, while every model bills output at roughly six times input. Culpa, a local-first LLM cost, margin, and forecast ledger, records which model served each call, so a drift up that ladder shows as a number rather than a surprise.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "150x", "claim": "spread between the cheapest and dearest OpenAI input rates", "confidence": "calculated", "source": "$30.00 per million on gpt-5.4-pro divided by $0.20 on gpt-5.4-nano, from the price book, effective 2026-07-02." }, { "value": "$0.45 to $66.00", "claim": "cost of one million input and 200,000 output tokens across the OpenAI line", "confidence": "calculated", "source": "Five tier calculations shown in the teardown, all from price-book rows effective 2026-07-02." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/openai-gpt-oss-120b-pricing", "url": "https://getculpa.com/openai-gpt-oss-120b-pricing", "textUrl": "https://getculpa.com/openai-gpt-oss-120b-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "gpt-oss-120b pricing", "title": "GPT-OSS-120B Pricing on Groq: Exactly Twice the 20B", "answer": "GPT-OSS-120B pricing on Groq is $0.15 per million input tokens and $0.60 per million output tokens, exactly double the 20B model at input, output and cached rates alike. That leaves no pricing subtlety to argue about, only whether the larger model earns twice the money. Culpa, a local-first LLM cost, margin, and forecast ledger, prices both against the same workload so the comparison is measured.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$0.15 / $0.60 per million", "claim": "GPT-OSS-120B input and output rates on Groq", "confidence": "calculated", "source": "Price book row for groq/openai/gpt-oss-120b, $0.00015 and $0.0006 per 1k, effective 2026-07-02, re-verified against Groq's rendered pricing page 2026-08-01." }, { "value": "2.00x", "claim": "cost of the 120B against the 20B, identical at input, output and cached rates", "confidence": "calculated", "source": "$0.15 divided by $0.075 input and $0.60 divided by $0.30 output, from both price-book rows." } ], "sources": [ { "label": "Groq pricing", "url": "https://groq.com/pricing" } ] }, { "slug": "/openai-gpt-oss-20b-pricing", "url": "https://getculpa.com/openai-gpt-oss-20b-pricing", "textUrl": "https://getculpa.com/openai-gpt-oss-20b-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "gpt-oss-20b pricing", "title": "GPT-OSS-20B Pricing on Groq: A Cache Worth Half, Not 90%", "answer": "GPT-OSS-20B pricing on Groq is $0.075 per million input tokens and $0.30 per million output tokens, with a cached input rate at about half the input price rather than the ninety percent discount hosted frontier models offer. Caching still pays here, just far less. Culpa, a local-first LLM cost, margin, and forecast ledger, prices cached and uncached input separately so the difference is visible.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$0.075 / $0.30 per million", "claim": "GPT-OSS-20B input and output rates on Groq", "confidence": "calculated", "source": "Price book row for groq/openai/gpt-oss-20b, $0.000075 and $0.0003 per 1k, effective 2026-07-02, re-verified against Groq's rendered pricing page 2026-08-01." }, { "value": "22% to 41%", "claim": "modelled saving on the same prefix at this model's cached discount against a ninety-percent one", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at real Groq rates. A range because prefix share and hit rate are modelled." } ], "sources": [ { "label": "Groq pricing", "url": "https://groq.com/pricing" } ] }, { "slug": "/openrouter-api-cost", "url": "https://getculpa.com/openrouter-api-cost", "textUrl": "https://getculpa.com/openrouter-api-cost/llms.txt", "pageType": "provider_cost", "intent": "EVALUATE", "primaryKeyword": "openrouter api cost", "title": "OpenRouter API Cost: One Model, Several Prices", "answer": "OpenRouter routes a single model to several upstream providers, and they don't all charge the same. Claude Opus 4.7 bills $5.00 or $5.50 per million input depending on which one served the call, a 10% spread you don't pick. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each call by the route that served it.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$5.00 / $25.00 per million", "claim": "OpenRouter Claude Opus 4.7 standard upstream rate, with a $0.50 cache read matching the direct route", "confidence": "calculated", "source": "OpenRouter's Claude Opus 4.7 model page, rendered and read 2026-08-01. Level with the anthropic/claude-opus-4-7 price book row at $0.005 and $0.025 per 1k, effective 2026-07-02." }, { "value": "$5,100 to $5,610", "claim": "modelled monthly cost of one repeated-prompt workload on the standard against the premium upstream provider", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at rates read from OpenRouter's Claude Opus 4.7 page on 2026-08-01. A range because call volume and cache hit rate are modelled." } ], "sources": [ { "label": "OpenRouter models", "url": "https://openrouter.ai/models" }, { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/openrouter-bill-spike", "url": "https://getculpa.com/openrouter-bill-spike", "textUrl": "https://getculpa.com/openrouter-bill-spike/llms.txt", "pageType": "problem", "intent": "AUDIT", "primaryKeyword": "openrouter bill spike", "title": "OpenRouter Bill Spike: Your Route Changed, Not Your Code", "answer": "An OpenRouter bill spike on flat traffic usually means a different upstream provider served your calls. The same model lists against several providers whose rates differ by about 10%, and routing follows availability rather than your intent. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each call by the route that served it, so a rate change shows up as a named cause.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$0.0170 to $0.0187", "claim": "modelled cost per call before and after an upstream route change, on identical traffic", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at rates read from OpenRouter's Claude Opus 4.7 page on 2026-08-01. A range because call volume and cache hit rate are modelled." }, { "value": "10%", "claim": "listed premium between upstream providers serving Claude Opus 4.7 on OpenRouter", "confidence": "calculated", "source": "$5.50 against $5.00 input and $27.50 against $25.00 output, read from OpenRouter's Claude Opus 4.7 model page on 2026-08-01." } ], "sources": [ { "label": "OpenRouter models", "url": "https://openrouter.ai/models" }, { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/output-token-creep", "url": "https://getculpa.com/output-token-creep", "textUrl": "https://getculpa.com/output-token-creep/llms.txt", "pageType": "problem", "intent": "FIX", "primaryKeyword": "output token creep", "title": "Output Token Creep: The Rise Input Monitoring Can't See", "answer": "Output token creep is a rise in tokens generated per call while input stays flat, so any monitoring built around prompt size sees nothing at all. It matters because output bills five to eight times input on most models here. Culpa, a local-first LLM cost, margin, and forecast ledger, tracks input and output separately per feature, so the two curves stop being reported as one number.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$0.001525 to $0.00265", "claim": "modelled cost per call before and after an output-length increase on Gemini 2.5 Flash", "confidence": "estimated", "source": "Both endpoints at real Gemini 2.5 Flash rates per 1k tokens from the price book, effective 2026-07-02. A range because the output lengths are modelled." }, { "value": "8.33x", "claim": "output-to-input price ratio on Gemini 2.5 Flash, the widest in the price book", "confidence": "calculated", "source": "$0.0025 output divided by $0.0003 input per 1k, from the price book, effective 2026-07-02." } ], "sources": [ { "label": "Gemini API pricing", "url": "https://ai.google.dev/gemini-api/docs/pricing" } ] }, { "slug": "/per-customer-llm-cost-forecast", "url": "https://getculpa.com/per-customer-llm-cost-forecast", "textUrl": "https://getculpa.com/per-customer-llm-cost-forecast/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "per customer llm cost forecast", "title": "Per-Customer LLM Cost Forecast: Forecasting the Total Hides the Risk", "answer": "A per-customer LLM cost forecast projects each account's spend from its own history rather than projecting one company-wide total. The aggregate can land perfectly while the customers inside it move in opposite directions. Culpa, a local-first LLM cost, margin, and forecast ledger, forecasts at the level it attributes, so a projection names an account.", "lastReviewed": "2026-08-03", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$900.00", "claim": "modelled company total, identical across two months while one account grew 75% and another fell 75%", "confidence": "calculated", "source": "Three modelled accounts priced on Claude Haiku 4.5 at real rates of $1.00 and $5.00 per million from the price book effective 2026-07-02. Month 1: $400.00 + $300.00 + $200.00 = $900.00. Month 2: $700.00 + $150.00 + $50.00 = $900.00. Account volumes are modelled and chosen so the total is flat, which is the entire point of the figure." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" }, { "label": "Datadog LLM Observability", "url": "https://www.datadoghq.com/product/llm-observability/" } ] }, { "slug": "/per-tier-token-budget", "url": "https://getculpa.com/per-tier-token-budget", "textUrl": "https://getculpa.com/per-tier-token-budget/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "per tier token budget", "title": "Per-Tier Token Budget: What a $29 Plan Can Actually Afford", "answer": "A subscription price is fixed and the usage behind it isn't, so every tier has a call count above which it loses money. Working it out takes one division, and skipping it explains how unlimited plans get priced. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each call and attributes it to the customer and plan behind it.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "527 calls", "claim": "modelled monthly calls a $29 tier can afford at an 80% target gross margin", "confidence": "calculated", "source": "A modelled 3,000 input and 500 output tokens per call on Claude Sonnet 5 at real introductory rates of $2.00 and $10.00 per million from the price book effective 2026-07-02 costs $0.0110. A modelled $29.00 tier at an 80% target gross margin leaves a $5.80 cost budget, and $5.80 / $0.0110 = 527 calls. A customer at 5,000 calls costs $55.00, a $26.00 loss against the $29.00 price. Rates are published, the price, margin target, call shape and usage are modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/pin-or-float-model-version", "url": "https://getculpa.com/pin-or-float-model-version", "textUrl": "https://getculpa.com/pin-or-float-model-version/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "pin or float model version", "title": "Pin or Float a Model Version: Pinning Can Cost You 3x", "answer": "Pinning a model version protects you from behaviour changing underneath you and exposes you to a different risk, because newer models are often cheaper. Anthropic's own list puts a deprecated model at three times its successor. Culpa, a local-first LLM cost, margin, and forecast ledger, records the model version on every call so the comparison is measurable.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "3x", "claim": "cost of a deprecated model against its current successor on identical traffic", "confidence": "calculated", "source": "Anthropic's pricing page at platform.claude.com/docs/en/docs/about-claude/pricing, read 2026-08-05, lists Claude Opus 4.1 as deprecated at $15.00 input and $75.00 output per million, and Claude Opus 5 at $5.00 and $25.00. A modelled 10 million input and 2 million output tokens costs $150.00 + $150.00 = $300.00 on the former and $50.00 + $50.00 = $100.00 on the latter, a factor of exactly 3 on both rates. The rates and deprecation markers are published, the volume is modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/poc-to-production-cost", "url": "https://getculpa.com/poc-to-production-cost", "textUrl": "https://getculpa.com/poc-to-production-cost/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "poc to production cost", "title": "POC to Production Cost: the One Number That Doesn't Survive", "answer": "A proof of concept measures cost per call under conditions production never repeats: short conversations, no retries, no tail users. Multiplying that figure by real volume under-forecasts, often badly. Culpa, a local-first LLM cost, margin, and forecast ledger, prices your real traffic so the forecast comes from production rather than from a demo.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$97,750.00 against $55,000.00", "claim": "modelled monthly cost at production shape against the same volume forecast from a pilot's cost per call", "confidence": "calculated", "source": "A modelled 3,000 input and 500 output tokens per call on Claude Sonnet 5 at real introductory rates of $2.00 and $10.00 per million from the price book effective 2026-07-02 gives $0.0110, so 5,000,000 calls forecasts $55,000.00. At a modelled production shape of 6,000 input, the same 500 output and a 15% retry rate, per-call cost is $0.0170 x 1.15 = $0.01955, so the same volume costs $97,750.00, which is $42,750.00 and 78% higher. Rates are published, every volume and shape is modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/power-user-ai-cost", "url": "https://getculpa.com/power-user-ai-cost", "textUrl": "https://getculpa.com/power-user-ai-cost/llms.txt", "pageType": "problem", "intent": "FIX", "primaryKeyword": "power user ai cost", "title": "Power User AI Cost: Why Your Heaviest Account Is Heavy", "answer": "Power user AI cost is what your heaviest accounts spend, and the useful question isn't how much but why. Most heavy accounts are heavy because of one workflow rather than general enthusiasm. Culpa, a local-first LLM cost, margin, and forecast ledger, attributes spend to the feature inside the account, so the answer names a workflow you can fix rather than a customer you'd rather not lose.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$240.00", "claim": "modelled monthly model spend for one heavy account across three workloads on GPT-5.4", "confidence": "calculated", "source": "$105.00 + $54.00 + $81.00 from the teardown arithmetic at real GPT-5.4 rates per 1k tokens from the price book, effective 2026-07-02. Workload volumes are modelled." }, { "value": "$86.10", "claim": "modelled share of that account's month spent re-processing documents that hadn't changed", "confidence": "calculated", "source": "82% of the $105.00 bulk re-run, from the teardown arithmetic at real GPT-5.4 rates. The unchanged share is modelled." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/price-a-credit-based-ai-plan", "url": "https://getculpa.com/price-a-credit-based-ai-plan", "textUrl": "https://getculpa.com/price-a-credit-based-ai-plan/llms.txt", "pageType": "margin", "intent": "EVALUATE", "primaryKeyword": "price a credit based ai plan", "title": "How to Price a Credit-Based AI Plan So a Credit Covers Cost", "answer": "A credit-based AI plan works only when one credit maps to a bounded amount of model spend. Teams usually set the credit price from what feels saleable, then let expensive actions cost the same credit as cheap ones. Culpa, a local-first LLM cost, margin, and forecast ledger, meters the real spend behind each action, so a credit gets priced from measurement.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$0.008 to $0.522", "claim": "modelled model spend behind a single credit, cheapest action against dearest", "confidence": "estimated", "source": "Both endpoints at real Claude Sonnet 5 introductory rates per 1k tokens from the price book, effective 2026-07-02. A range because the action token volumes are modelled." }, { "value": "$0.049", "claim": "revenue per credit on a modelled $49 plan carrying 1,000 credits", "confidence": "calculated", "source": "$49 divided by 1,000 credits. The plan shape is modelled, the rates behind each action are real." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/price-a-per-seat-ai-product", "url": "https://getculpa.com/price-a-per-seat-ai-product", "textUrl": "https://getculpa.com/price-a-per-seat-ai-product/llms.txt", "pageType": "margin", "intent": "EVALUATE", "primaryKeyword": "price a per seat ai product", "title": "How to Price a Per-Seat AI Product When Seats Differ", "answer": "Per-seat pricing decouples revenue from usage, which is comfortable until seats stop looking alike. On AI products an active seat can cost hundreds of times a dormant one, so the seats you bill and the seats you serve drift apart. Culpa, a local-first LLM cost, margin, and forecast ledger, attributes spend to each seat, so the drift becomes a number instead of a surprise.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$0.162 to $63.60", "claim": "modelled monthly model spend for one seat, dormant against power user, on a $40 seat price", "confidence": "estimated", "source": "Both endpoints at real Gemini 3.5 Flash rates per 1k tokens from the price book, effective 2026-07-02. A range because the seat mix and token volumes are modelled." }, { "value": "86.3%", "claim": "modelled blended gross margin across 1,000 seats, of which 380 are dormant", "confidence": "calculated", "source": "($40,000 minus $5,464.56) divided by $40,000, from the teardown arithmetic at real Gemini 3.5 Flash rates." } ], "sources": [ { "label": "Gemini API pricing", "url": "https://ai.google.dev/gemini-api/docs/pricing" } ] }, { "slug": "/price-a-usage-based-ai-product", "url": "https://getculpa.com/price-a-usage-based-ai-product", "textUrl": "https://getculpa.com/price-a-usage-based-ai-product/llms.txt", "pageType": "margin", "intent": "EVALUATE", "primaryKeyword": "price a usage based ai product", "title": "How to Price a Usage-Based AI Product on a Real Markup", "answer": "Usage-based pricing passes model cost through to the customer, so the only real decision is the markup and what it covers. A markup set on accepted spend under-recovers, because retries, evaluation traffic and calls that produced nothing you kept all bill too. Culpa, a local-first LLM cost, margin, and forecast ledger, counts every billed call, so the markup sits on real spend.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$57.00 to $71.18", "claim": "modelled monthly provider spend for one account, accepted work alone against everything billed", "confidence": "estimated", "source": "Both endpoints at real GPT-5.4 mini rates per million from the price book, effective 2026-07-02. A range because the waste and evaluation shares are modelled." }, { "value": "37.6%", "claim": "modelled gross margin when a 2x markup is applied to accepted spend rather than billed spend", "confidence": "calculated", "source": "($114.00 minus $71.18) divided by $114.00, from the teardown arithmetic at real GPT-5.4 mini rates." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/price-an-unlimited-ai-plan", "url": "https://getculpa.com/price-an-unlimited-ai-plan", "textUrl": "https://getculpa.com/price-an-unlimited-ai-plan/llms.txt", "pageType": "margin", "intent": "EVALUATE", "primaryKeyword": "price an unlimited ai plan", "title": "How to Price an Unlimited AI Plan Without Going Broke", "answer": "An unlimited AI plan prices a promise against a maximum nobody has computed, and on a long-tailed usage curve that maximum keeps moving. A generous included allowance with a stated overage reads nearly as well and bounds the risk. Culpa, a local-first LLM cost, margin, and forecast ledger, reports the spend distribution so the tail is a number rather than a fear.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$40 to $850", "claim": "modelled monthly cost range on a single $200 unlimited tier", "confidence": "estimated", "source": "Both endpoints at real Claude Haiku 4.5 rates per 1k tokens from the price book, effective 2026-07-02. A range because the usage profiles are modelled." }, { "value": "-325% to 80%", "claim": "modelled gross margin spread from typical to extreme account on a $200 unlimited tier", "confidence": "estimated", "source": "($200 - $850) / $200, from the teardown arithmetic at real Claude Haiku 4.5 rates. Modelled usage, so published as a scenario." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/prompt-cache-cost-calculator", "url": "https://getculpa.com/prompt-cache-cost-calculator", "textUrl": "https://getculpa.com/prompt-cache-cost-calculator/llms.txt", "pageType": "calculator", "intent": "HOW_TO", "primaryKeyword": "prompt cache cost calculator", "title": "Prompt Cache Cost Calculator: Does Caching Pay on Your Traffic", "answer": "This prompt cache cost calculator takes your shared prefix size and the number of calls that land inside one cache window, then prices the cached path against the uncached one. Culpa, a local-first LLM cost, margin, and forecast ledger, computes both in exact decimal, so you can see whether the write premium earns itself back on your own traffic.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": "cache", "costFigures": [ { "value": "2 calls", "claim": "calls inside one five-minute window at which caching starts to pay", "confidence": "calculated", "source": "A write costs 1.25 times input and a read 0.10 times, so M = 1.25 + (M - 1) x 0.10 solves to M = 1.28. Anthropic's multipliers from the price book, effective 2026-07-02." }, { "value": "$0.1350 to $1.0000", "claim": "modelled monthly cost of one prefix across the teardown's arrival patterns, cheapest to dearest", "confidence": "estimated", "source": "Endpoints from the teardown arithmetic at a real Claude Sonnet 5 rate. A range because the arrival patterns are modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/prompt-cache-cost-optimization", "url": "https://getculpa.com/prompt-cache-cost-optimization", "textUrl": "https://getculpa.com/prompt-cache-cost-optimization/llms.txt", "pageType": "optimization", "intent": "IMPROVE", "primaryKeyword": "prompt cache cost optimization", "title": "Prompt Cache Cost Optimization: The Second Call Decides It", "answer": "Caching pays from the second call that lands inside the cache window, because a five-minute write bills at 1.25 times the input rate while a read bills at a tenth. A one-hour write needs a third call. Culpa, a local-first LLM cost, margin, and forecast ledger, measures how many of your calls arrive inside a window rather than how many arrive in a day.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "2 calls", "claim": "calls inside one five-minute window at which caching starts to pay", "confidence": "calculated", "source": "M = 1.25 + (M - 1) x 0.10 solves to M = 1.28, using Anthropic's published write and read multipliers from the price book, effective 2026-07-02 and re-verified 2026-08-02." }, { "value": "3 calls", "claim": "the same break-even once the one-hour cache doubles the write premium", "confidence": "calculated", "source": "M = 2.00 + (M - 1) x 0.10 solves to M = 2.11, using the same published multipliers." }, { "value": "$0.315 to $2.50", "claim": "modelled monthly cost of twenty calls on one prefix, best arrival pattern to worst", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at a real Claude Sonnet 5 rate. A range because the arrival patterns are modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/prompt-cache-cost-teardown", "url": "https://getculpa.com/prompt-cache-cost-teardown", "textUrl": "https://getculpa.com/prompt-cache-cost-teardown/llms.txt", "pageType": "teardown", "intent": "EXAMPLES", "primaryKeyword": "prompt cache cost teardown", "title": "Prompt Cache Cost Teardown: The Month Caching Cost Money", "answer": "This is a real month of Anthropic API usage, measured rather than modelled. The account read back 0.2035 tokens for every one it wrote into a prompt cache, against a break-even of 0.28. Culpa, a local-first LLM cost, margin, and forecast ledger, meters cache writes and reads separately, which is the only way that gap becomes visible.", "lastReviewed": "2026-08-03", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "27.2%", "claim": "share of one real month's Anthropic API bill spent creating prompt caches", "confidence": "provider-reported", "source": "Anthropic usage export for the month, input_cache_write_5m line item, aggregated by scripts/analyse-usage-export.py. Account and workspace identifiers stripped on load." }, { "value": "0.45%", "claim": "share of the same month's bill spent reading those caches back", "confidence": "provider-reported", "source": "Anthropic usage export for the same month, input_cache_read line item, from the same aggregation. The two line items stand at about 60 to 1." }, { "value": "0.204", "claim": "measured cache reads per cache write, against a break-even of 0.28", "confidence": "calculated", "source": "Cache-read tokens divided by cache-write tokens, both from the provider's usage export, published as a ratio because the counts themselves would recover a withheld total." }, { "value": "5.55%", "claim": "how much more the month cost with caching than the same tokens would have cost without it", "confidence": "calculated", "source": "(1.25 + (0.2035 x 0.10)) divided by (1 + 0.2035), using Anthropic's published cache multipliers on the export's own reads-per-write ratio. Carried to four figures because rounding to 0.204 first gives 5.51% and no longer reproduces the published number." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/prompt-cache-miss-cost", "url": "https://getculpa.com/prompt-cache-miss-cost", "textUrl": "https://getculpa.com/prompt-cache-miss-cost/llms.txt", "pageType": "problem", "intent": "FIX", "primaryKeyword": "prompt cache miss cost", "title": "Prompt Cache Miss Cost: What a Broken Cache Really Costs", "answer": "A prompt cache miss bills your repeated prompt at the full input rate instead of the cached rate, which on some models is ten times more. Culpa, a local-first LLM cost, margin, and forecast ledger, meters cached, uncached and cache-write tokens separately per call, so a cache that stopped working shows up the day it breaks.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "10x", "claim": "cost of a cache miss against a cache hit on Claude Haiku 4.5", "confidence": "calculated", "source": "$0.001 input rate divided by the $0.0001 cached rate, both per 1k tokens from the Claude Haiku 4.5 price book row, effective 2026-07-02." }, { "value": "$400 to $4,000", "claim": "modelled monthly cost of a 4,000-token system prompt at one million calls, fully cached versus fully missed", "confidence": "estimated", "source": "Endpoints from the teardown arithmetic at real Claude Haiku 4.5 rates. A range because it brackets the two extremes of hit rate." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/prompt-version-cost-increase", "url": "https://getculpa.com/prompt-version-cost-increase", "textUrl": "https://getculpa.com/prompt-version-cost-increase/llms.txt", "pageType": "problem", "intent": "FIX", "primaryKeyword": "prompt version cost increase", "title": "Prompt Version Cost Increase: Finding the Edit Behind It", "answer": "A prompt version cost increase shows up as a step change in cost per call on one specific date, with call volume unchanged. Finding the edit behind it needs a prompt version recorded on each call, because an invoice can't be bisected after the fact. Culpa, a local-first LLM cost, margin, and forecast ledger, names the prompt version behind every call, so the step resolves to a diff.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$0.0038182 to $0.0058636", "claim": "modelled cost per call either side of a single-day step change on GPT-5.4", "confidence": "estimated", "source": "Daily spend divided by daily calls, at real GPT-5.4 rates per 1k tokens from the price book, effective 2026-07-02. A range because the daily figures are modelled." }, { "value": "818 tokens", "claim": "input tokens added per call, derived from the observed per-call cost delta", "confidence": "calculated", "source": "$0.0020455 divided by the $0.0025 per 1k input rate, with output length held constant." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/provider-switching-cost", "url": "https://getculpa.com/provider-switching-cost", "textUrl": "https://getculpa.com/provider-switching-cost/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "provider switching cost", "title": "Provider Switching Cost: Your Token Counts Don't Come With You", "answer": "Comparing rate cards is the easy part of a migration and the smallest part of the cost. The same text produces different token counts on different tokenizers, so your volumes change along with your rates. Culpa, a local-first LLM cost, margin, and forecast ledger, prices your real traffic against a versioned price book so both sides move together.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$260.00 against $200.00", "claim": "modelled monthly cost of identical text before and after a published tokenizer change", "confidence": "calculated", "source": "Anthropic's pricing page states that Claude 4.7 and later use a newer tokenizer producing approximately 30% more tokens for the same text, with Claude Sonnet 4.6 and earlier on the previous one. A modelled 100 million input tokens becomes approximately 130 million for identical text, and at Claude Sonnet 5's real introductory rate of $2.00 per million from the price book effective 2026-07-02 gives $200.00 against $260.00. The tokenizer statement and the rate are published, the volume is modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/qwen-qwen3-6-27b-pricing", "url": "https://getculpa.com/qwen-qwen3-6-27b-pricing", "textUrl": "https://getculpa.com/qwen-qwen3-6-27b-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "qwen3.6 27b pricing", "title": "Qwen3.6 27B Pricing on Groq: Five-to-One Output, No Cache", "answer": "Qwen3.6 27B pricing on Groq is $0.60 per million input tokens and $3.00 per million output tokens, a five-to-one spread, and the price book records no cached rate for it. A model tuned to reason out loud on a five-times output spread with no cache is the most expensive combination here. Culpa, a local-first LLM cost, margin, and forecast ledger, meters output tokens per feature so the reasoning is priced.", "lastReviewed": "2026-08-01", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$0.60 / $3.00 per million", "claim": "Qwen3.6 27B input and output rates on Groq", "confidence": "calculated", "source": "Price book row for groq/qwen/qwen3.6-27b, $0.0006 and $0.003 per 1k, effective 2026-07-02, re-verified against Groq's rendered pricing page 2026-08-01." }, { "value": "5.00x", "claim": "output rate as a multiple of input, the widest of any Groq model in this price book", "confidence": "calculated", "source": "$3.00 divided by $0.60, from the same price-book row." }, { "value": "$19.20 to $31.20", "claim": "modelled monthly cost with reasoning off and on, at identical input volume", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at real Groq rates. A range because the token volumes are modelled." } ], "sources": [ { "label": "Groq pricing", "url": "https://groq.com/pricing" } ] }, { "slug": "/rag-context-cost", "url": "https://getculpa.com/rag-context-cost", "textUrl": "https://getculpa.com/rag-context-cost/llms.txt", "pageType": "problem", "intent": "FIX", "primaryKeyword": "rag context cost", "title": "RAG Context Cost: Top-K Is a Pricing Decision", "answer": "RAG context cost is the share of your bill created by retrieved chunks rather than by the question a user asked. Top-k multiplied by chunk size sets input tokens per call, so retrieval settings are a pricing decision made in a config file. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each call and maps the input back to the feature that assembled it.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$0.0102 to $0.0186", "claim": "modelled cost per call at top-k 5 against top-k 12 on Gemini 3.5 Flash", "confidence": "estimated", "source": "Both endpoints at real Gemini 3.5 Flash rates per 1k tokens from the price book, effective 2026-07-02. A range because chunk size and answer length are modelled." }, { "value": "93%", "claim": "share of input tokens that are retrieved context at top-k 12 with 800-token chunks", "confidence": "calculated", "source": "9,600 retrieved tokens divided by 10,300 total input tokens, from the teardown arithmetic." } ], "sources": [ { "label": "Gemini API pricing", "url": "https://ai.google.dev/gemini-api/docs/pricing" } ] }, { "slug": "/selling-ai-credits-margin", "url": "https://getculpa.com/selling-ai-credits-margin", "textUrl": "https://getculpa.com/selling-ai-credits-margin/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "selling ai credits margin", "title": "Selling AI Credits: One Price, Ten Times the Cost", "answer": "Selling AI credits sets a fixed price against a variable cost, because the customer decides how much work each credit does. Margin per credit is a range, not a number, and only per-call pricing reveals its shape. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each redemption and attributes it to the customer who spent it.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "94.5% against 45.0%", "claim": "modelled margin on the same $0.10 credit, redeemed lightly and heavily", "confidence": "calculated", "source": "A light redemption of 3,000 input and 500 output tokens on Claude Haiku 4.5 at real rates of $1.00 and $5.00 per million from the price book effective 2026-07-02 costs $0.0030 + $0.0025 = $0.0055, giving ($0.10 - $0.0055) / $0.10 = 94.5%. A heavy redemption of 30,000 and 5,000 costs $0.0300 + $0.0250 = $0.0550, giving 45.0%. The credit price, both token shapes and the redemption behaviour are modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/shadow-ai-spend", "url": "https://getculpa.com/shadow-ai-spend", "textUrl": "https://getculpa.com/shadow-ai-spend/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "shadow ai spend", "title": "Shadow AI Spend: the Tools You're Paying For and Can't See", "answer": "Shadow AI spend is money leaving the company on tools nobody registered, usually through personal subscriptions and expense claims rather than the API budget everyone watches. It's a procurement problem before it's a cost one. Culpa, a local-first LLM cost, margin, and forecast ledger, prices the traffic you do control so the gap is visible.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$9,600.00 a year", "claim": "modelled annual cost of 40 personal AI subscriptions expensed individually", "confidence": "calculated", "source": "A modelled 40 staff each holding one $20 monthly personal subscription: 40 x $20 = $800.00 a month, and $800.00 x 12 = $9,600.00 a year. Every input is modelled and deliberately conservative, assuming one subscription each with no team tools, bundled add-ons or externally held API keys." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/speech-to-text-cost", "url": "https://getculpa.com/speech-to-text-cost", "textUrl": "https://getculpa.com/speech-to-text-cost/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "speech to text cost", "title": "Speech-to-Text Cost: Priced by the Minute, Not the Word", "answer": "Transcription and realtime audio bill per minute of audio rather than per token, so cost tracks how long the recording is and not how much was said. Rates differ elevenfold across models. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each call from a versioned price book so a model change is visible immediately.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$18,600.00 per month", "claim": "modelled difference between the cheapest and dearest transcription models on identical audio", "confidence": "calculated", "source": "OpenAI's pricing page, read 2026-08-05, publishes gpt-4o-mini-transcribe at $0.003 per minute and gpt-realtime-translate at $0.034 per minute. At a modelled 10,000 hours a month, which is 600,000 minutes, gives $1,800.00 against $20,400.00, a difference of $18,600.00 and a ratio of 11.33. Whisper at $0.006 gives $3,600.00. Rates are published, the volume is modelled." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" }, { "label": "Gemini API pricing", "url": "https://ai.google.dev/gemini-api/docs/pricing" } ] }, { "slug": "/stream-interruption-cost", "url": "https://getculpa.com/stream-interruption-cost", "textUrl": "https://getculpa.com/stream-interruption-cost/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "stream interruption cost", "title": "Stream Interruption Cost: You Pay Twice to Finish Once", "answer": "An interrupted stream still bills for everything generated before it stopped, and the documented way to resume re-sends that partial output as input. So you pay to generate it once and to read it back again. Culpa, a local-first LLM cost, margin, and forecast ledger, prices both halves as the two separate calls they really are.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "23.6%", "claim": "modelled overhead of one interruption plus resume against a clean run", "confidence": "calculated", "source": "A modelled 2,000-token prompt and 4,000-token answer on Claude Sonnet 5 at real rates of $2.00 and $10.00 per million from the price book effective 2026-07-02 costs $0.044000 clean. An interruption at 80% of output costs $0.036000 with no answer, and a resume carrying the 3,200 partial tokens forward as input costs $0.010400 + $0.008000 = $0.018400, totalling $0.054400, which is $0.010400 more or 23.6%. Rates and the documented recovery method are published, the token counts and interruption point are modelled." } ], "sources": [ { "label": "Anthropic, streaming messages", "url": "https://platform.claude.com/docs/en/build-with-claude/streaming" }, { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/tencent-hy3-preview-pricing", "url": "https://getculpa.com/tencent-hy3-preview-pricing", "textUrl": "https://getculpa.com/tencent-hy3-preview-pricing/llms.txt", "pageType": "model_pricing", "intent": "PRICE", "primaryKeyword": "tencent hy3 preview pricing", "title": "Tencent HY3 Preview Pricing: What a Preview Rate Is Worth", "answer": "Tencent HY3 Preview bills $0.063 per million input tokens and $0.21 output through OpenRouter, with cached input at $0.021. It's a preview model, so that rate carries no commitment to stay. Culpa, a local-first LLM cost, margin, and forecast ledger, dates every rate it stores so a preview repricing shows up as a change rather than as noise.", "lastReviewed": "2026-08-03", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$0.063 per million", "claim": "Tencent HY3 Preview input rate through OpenRouter", "confidence": "calculated", "source": "openrouter.ai/api/v1/models reports 0.000000063 per input token, read 2026-08-03. Multiplied to a per-million figure, and unchanged from Culpa's 2026-07-02 book row." }, { "value": "$52.50 to $98.00", "claim": "modelled monthly cost of one workload on the preview rate and on the nearest alternative", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at real rates. A range because the token volumes are modelled." } ], "sources": [ { "label": "OpenRouter models", "url": "https://openrouter.ai/models" }, { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/thinking-token-cost", "url": "https://getculpa.com/thinking-token-cost", "textUrl": "https://getculpa.com/thinking-token-cost/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "thinking token cost", "title": "Thinking Token Cost: You Pay for Reasoning You Never See", "answer": "Thinking tokens bill as output tokens whether or not the reasoning is shown to you, so the visible text understates what you paid. A separate usage field reports the real figure. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each call from the totals the provider reports rather than from what came back visible.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "12.5x", "claim": "cost of a cache write against the cache read it replaces when an effort change invalidates the cache", "confidence": "calculated", "source": "Anthropic's pricing page publishes cache multipliers of 1.25x base input for a 5-minute write and 0.1x for a read, which on Claude Opus 4.8 is $6.25 and $0.50 per million against $5.00 base. $6.25 / $0.50 = 12.5. Applied to the 3,546 cached tokens in Anthropic's own published effort-change trace, the write costs $0.022163 against the read's $0.001773, a difference of $0.020390. Both the multipliers and the token count are published." } ], "sources": [ { "label": "Anthropic, steering thinking and its cost", "url": "https://platform.claude.com/docs/en/build-with-claude/thinking-steering-and-cost" }, { "label": "Anthropic, thinking", "url": "https://platform.claude.com/docs/en/build-with-claude/thinking" }, { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/three-month-cost-teardown", "url": "https://getculpa.com/three-month-cost-teardown", "textUrl": "https://getculpa.com/three-month-cost-teardown/llms.txt", "pageType": "teardown", "intent": "EXAMPLES", "primaryKeyword": "llm usage trend three months", "title": "Three Months of Real LLM Usage: The Same Account, Three Products", "answer": "Three consecutive months of one account's Anthropic token usage, measured. Input volume per active day ran 100, then 199, then 18 against the first month, and caching appeared in month two and was gone by month three. Culpa, a local-first LLM cost, margin, and forecast ledger, keeps the history that makes a change that size legible instead of surprising.", "lastReviewed": "2026-08-03", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "100, 199, 18", "claim": "input volume per active day across three consecutive months, indexed on the first", "confidence": "calculated", "source": "Anthropic API token exports for 2026-05, 2026-06 and 2026-07, summed across all four input token types and divided by the active days in each export, then indexed on May. Aggregated by scripts/analyse-usage-export.py, which drops workspace, key and account identifiers on load." }, { "value": "0.34, 0.19, 0.25", "claim": "output tokens as a share of input tokens, month by month", "confidence": "calculated", "source": "Output tokens divided by total input tokens for each month, from the same exports. Published as ratios because the counts would recover a withheld total." }, { "value": "0% to 19.1% to 0%", "claim": "cache-write tokens as a share of all input, across the three months", "confidence": "calculated", "source": "The cache-write token fields divided by total input tokens, straight from the provider's own export for each month." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/tool-definition-token-cost", "url": "https://getculpa.com/tool-definition-token-cost", "textUrl": "https://getculpa.com/tool-definition-token-cost/llms.txt", "pageType": "problem", "intent": "FIX", "primaryKeyword": "tool definition token cost", "title": "Tool Definition Token Cost: You Pay for Every Unused Tool", "answer": "Tool definitions are input tokens, so every schema you attach bills on every call whether the model uses it or not. A large tool block is a fixed tax on all traffic. Culpa, a local-first LLM cost, margin, and forecast ledger, meters input tokens per call, so a growing tool block shows up as a rising floor rather than a mystery.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$0.0064 per call", "claim": "cost of a 3,200-token tool definition block at real Claude Sonnet 5 input rates", "confidence": "calculated", "source": "3.2 x $0.002 per 1k input tokens, using the Claude Sonnet 5 introductory rate from the price book, effective 2026-07-02." }, { "value": "$9,000 to $32,000", "claim": "modelled monthly tool-block spend at five million calls, trimmed versus untrimmed", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at the real Claude Sonnet 5 input rate. A range because call volume and block size are modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/track-llm-cost-anthropic", "url": "https://getculpa.com/track-llm-cost-anthropic", "textUrl": "https://getculpa.com/track-llm-cost-anthropic/llms.txt", "pageType": "integration", "intent": "HOW_TO", "primaryKeyword": "track llm cost anthropic", "title": "Track LLM Cost on Anthropic, Where Usage Arrives in Two Parts", "answer": "Anthropic splits streamed usage across two events, one carrying input and a later one carrying output, and it charges to create a cache as well as to read one. Culpa, a local-first LLM cost, margin, and forecast ledger, meters both cache writes and cache reads as separate quantities, and takes the cheaper reading whenever the response leaves the cache duration unstated.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$125.00 to $200.00", "claim": "modelled monthly cost of 20M cache-creation tokens at the five-minute against the one-hour write rate", "confidence": "estimated", "source": "Both endpoints at real Claude Opus 4.8 cache-write rates from the price book, effective 2026-07-02. A range because the cache-creation volume is modelled." }, { "value": "$75.00", "claim": "modelled exposure absorbed by taking the cheaper write rate when the duration is unstated", "confidence": "calculated", "source": "$200.00 minus $125.00, from the teardown arithmetic at real price-book rates." } ], "sources": [ { "label": "Culpa docs: connect your providers", "url": "https://app.getculpa.com/docs/how-to/connect-your-providers" }, { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/track-llm-cost-by-agent-step", "url": "https://getculpa.com/track-llm-cost-by-agent-step", "textUrl": "https://getculpa.com/track-llm-cost-by-agent-step/llms.txt", "pageType": "attribution", "intent": "HOW_TO", "primaryKeyword": "track llm cost by agent step", "title": "Track LLM Cost by Agent Step: The Join Is a Trace ID", "answer": "To track LLM cost by agent step you carry a trace identifier through every call the run makes, then sum the priced calls back to that identifier. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each call from a versioned price book and attributes it to the step and the run that produced it, so a run total exists rather than a pile of calls.", "lastReviewed": "2026-08-03", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$0.0516", "claim": "modelled cost of one seven-call agent run on Claude Haiku 4.5", "confidence": "calculated", "source": "Arithmetic shown in full on this page, using real rates of $1.00 and $5.00 per million from the price book effective 2026-07-02. The seven-step shape and every token count are modelled, not measured. The figure is on the page to show the spread across steps rather than to claim a typical run cost." }, { "value": "63%", "claim": "share of that modelled run carried by a single retrieval summarisation step", "confidence": "calculated", "source": "$0.0325 of a $0.0516 run total = 63.0%, from the same modelled step shape and the same real rates. Reported as a share because the share is the transferable finding: one call in seven can dominate a run while looking ordinary in a call count." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/track-llm-cost-by-api-key", "url": "https://getculpa.com/track-llm-cost-by-api-key", "textUrl": "https://getculpa.com/track-llm-cost-by-api-key/llms.txt", "pageType": "attribution", "intent": "HOW_TO", "primaryKeyword": "track llm cost by api key", "title": "Track LLM Cost by API Key: Cheapest to Build, First to Mislead", "answer": "Tracking LLM cost by API key means issuing a separate provider key per tenant or environment and reading spend per key. It needs no code change, which is why most teams start there. Culpa, a local-first LLM cost, margin, and forecast ledger, prices and attributes each call individually, so attribution survives a key that serves many customers.", "lastReviewed": "2026-08-03", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$890.00", "claim": "modelled monthly spend on one shared key, hiding a 7x spread between the customers behind it", "confidence": "calculated", "source": "Sum of three modelled customers priced on gpt-5.4 at real rates of $2.50 and $15.00 per million from the price book effective 2026-07-02: $480.00 + $345.00 + $65.00. Customer token volumes are modelled. The point of the figure is that the total is correct and the three components are unavailable from it." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/track-llm-cost-by-prompt-version", "url": "https://getculpa.com/track-llm-cost-by-prompt-version", "textUrl": "https://getculpa.com/track-llm-cost-by-prompt-version/llms.txt", "pageType": "attribution", "intent": "HOW_TO", "primaryKeyword": "track llm cost by prompt version", "title": "Track LLM Cost by Prompt Version: Or the Regression Is Invisible", "answer": "To track LLM cost by prompt version you record which version produced each call, then price the calls and compare versions on cost per call rather than on total spend. Culpa, a local-first LLM cost, margin, and forecast ledger, tags every call with its version and prices it from a versioned price book, so a prompt change and a rate change stay separable.", "lastReviewed": "2026-08-03", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$720.00 per month", "claim": "modelled cost of adding 180 input tokens to a system prompt at 4M calls a month", "confidence": "calculated", "source": "180 tokens x 4,000,000 calls = 720,000,000 input tokens, at the real Claude Haiku 4.5 input rate of $1.00 per million from the price book effective 2026-07-02, giving $720.00. The token increment and the call volume are modelled. The rate is real." }, { "value": "$0.00018", "claim": "the same change expressed per call, where it disappears into the noise", "confidence": "calculated", "source": "180 tokens at $1.00 per million = $0.00018 per call, from the same real rate. Shown beside the monthly figure because the pair is the finding: the same change is invisible at one granularity and material at another." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/track-llm-cost-by-tool-call", "url": "https://getculpa.com/track-llm-cost-by-tool-call", "textUrl": "https://getculpa.com/track-llm-cost-by-tool-call/llms.txt", "pageType": "attribution", "intent": "HOW_TO", "primaryKeyword": "track llm cost by tool call", "title": "Track LLM Cost by Tool Call: The Second Bill Nobody Attributes", "answer": "To track LLM cost by tool call you record which tool a call belongs to, including the model calls tools make internally, then price and roll them up to the request that triggered them. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each call and attributes it to the tool and the run, so nested spend lands where it was caused.", "lastReviewed": "2026-08-03", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "$7,300.00 per month", "claim": "modelled cost attributable to one tool, counting its schema tax and its nested model calls", "confidence": "calculated", "source": "Arithmetic shown in full on this page. 1,400 schema tokens x 2,000,000 requests = $2,800.00, plus 300,000 invocations x (12,000 input and 600 output) = $3,600.00 + $900.00 = $4,500.00. Real Claude Haiku 4.5 rates of $1.00 and $5.00 per million from the price book effective 2026-07-02. Every volume, the schema size and the invocation rate are modelled." }, { "value": "$2,800.00 per month", "claim": "the schema portion alone, charged on requests that invoke no tool at all", "confidence": "calculated", "source": "1,400 tool-definition tokens x 2,000,000 requests = 2,800,000,000 input tokens, at the real $1.00 per million input rate from the price book effective 2026-07-02. Reported separately because it scales on total requests rather than on invocations, which is why an invocation-count dashboard never shows it." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/track-llm-cost-gemini", "url": "https://getculpa.com/track-llm-cost-gemini", "textUrl": "https://getculpa.com/track-llm-cost-gemini/llms.txt", "pageType": "integration", "intent": "HOW_TO", "primaryKeyword": "track llm cost gemini", "title": "Track LLM Cost on Gemini: Two Doors, Different Numbers", "answer": "Gemini can be reached two ways, through its native API or through an OpenAI-compatible endpoint, and the two carry usage in different fields. Culpa, a local-first LLM cost, margin, and forecast ledger, reads whichever shape arrives, and the route you chose decides which counts are available to price, thinking tokens most of all.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$52.50 to $112.50", "claim": "modelled monthly Gemini cost recorded from visible output alone against the full billed total", "confidence": "estimated", "source": "Both endpoints at real Gemini 2.5 Pro rates per million from the price book, effective 2026-07-02. A range because the token volumes and the thinking share are modelled." }, { "value": "53%", "claim": "modelled share of a reasoning workload's bill sitting in tokens no user ever sees", "confidence": "calculated", "source": "$60.00 of thinking-token output divided by the $112.50 modelled total, from the teardown arithmetic." } ], "sources": [ { "label": "Culpa docs: connect your providers", "url": "https://app.getculpa.com/docs/how-to/connect-your-providers" }, { "label": "Gemini API pricing", "url": "https://ai.google.dev/gemini-api/docs/pricing" } ] }, { "slug": "/track-llm-cost-groq", "url": "https://getculpa.com/track-llm-cost-groq", "textUrl": "https://getculpa.com/track-llm-cost-groq/llms.txt", "pageType": "integration", "intent": "HOW_TO", "primaryKeyword": "track llm cost groq", "title": "Track LLM Cost on Groq When a Model Has No Price Yet", "answer": "Open-weight lineups change faster than any price book tracks, so a Groq model can arrive or vanish between updates. Culpa, a local-first LLM cost, margin, and forecast ledger, prices a call it has no rate for at zero and flags it as incomplete, so an unpriced model shows up as a visible gap rather than as a silently wrong total.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$1,196.80", "claim": "modelled monthly spend recoverable from stored counts once a missing rate is added", "confidence": "calculated", "source": "800,000 calls x $0.001496, using real Groq Llama 3.3 70B rates per million from the price book, effective 2026-07-02 and re-verified 2026-08-01. Call volume and token counts are modelled." }, { "value": "$0.00", "claim": "what an unpriced model contributes to a total before its rate exists, alongside an incomplete flag", "confidence": "calculated", "source": "A model with no price-book row prices to zero by construction, and the completeness flag rather than the amount carries the meaning." } ], "sources": [ { "label": "Culpa docs: connect your providers", "url": "https://app.getculpa.com/docs/how-to/connect-your-providers" }, { "label": "Groq pricing", "url": "https://groq.com/pricing" } ] }, { "slug": "/track-llm-cost-langchain", "url": "https://getculpa.com/track-llm-cost-langchain", "textUrl": "https://getculpa.com/track-llm-cost-langchain/llms.txt", "pageType": "integration", "intent": "HOW_TO", "primaryKeyword": "track llm cost langchain", "title": "Track LLM Cost in LangChain, and What the Adapter Can See", "answer": "Culpa's LangChain adapter is a callback handler you attach to your calls, and it reports the token counts LangChain itself surfaces. That's an honest floor rather than a ceiling, because the adapter sees what the framework reports and not the raw provider response. Culpa, a local-first LLM cost, margin, and forecast ledger, prices what arrives and reconciles it against the invoice so the floor is measurable.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$215 to $350", "claim": "modelled monthly cost of one workload, priced with the cached split and priced without it", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at real Claude Opus 4.8 rates from the price book, effective 2026-07-02. A range because the cached share is modelled." }, { "value": "63%", "claim": "modelled over-statement when a cached-token field never reaches the ledger", "confidence": "calculated", "source": "$350 divided by $215, from the teardown arithmetic." } ], "sources": [ { "label": "langchain-core API reference", "url": "https://reference.langchain.com/python/langchain-core" }, { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/track-llm-cost-litellm", "url": "https://getculpa.com/track-llm-cost-litellm", "textUrl": "https://getculpa.com/track-llm-cost-litellm/llms.txt", "pageType": "integration", "intent": "HOW_TO", "primaryKeyword": "track llm cost litellm", "title": "Track LLM Cost in LiteLLM: The Highest-Fidelity Adapter", "answer": "Culpa's LiteLLM adapter is a logger you register on litellm.callbacks, and it carries more of the truth than any other framework path. LiteLLM's own response object is already openai-shaped, so it passes through nearly verbatim rather than being summarised. Culpa, a local-first LLM cost, margin, and forecast ledger, prices that response the same way it prices a call captured on the request path.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$900 to $1,024.25", "claim": "modelled monthly spend at the start and end of six months of imported history", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at real GPT-5.4 rates from the price book, effective 2026-07-02. A range because the volumes and the growth rate are modelled." }, { "value": "27.6%", "claim": "modelled input growth across six months with output volume unchanged", "confidence": "calculated", "source": "1.05 to the power of 5 equals 1.2763, from the teardown arithmetic." } ], "sources": [ { "label": "LiteLLM custom callbacks", "url": "https://docs.litellm.ai/docs/observability/custom_callback" }, { "label": "Culpa docs: import a usage file", "url": "https://app.getculpa.com/docs/how-to/import-a-usage-file" }, { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/track-llm-cost-ollama", "url": "https://getculpa.com/track-llm-cost-ollama", "textUrl": "https://getculpa.com/track-llm-cost-ollama/llms.txt", "pageType": "integration", "intent": "HOW_TO", "primaryKeyword": "track llm cost ollama", "title": "Track LLM Cost for Ollama, Where No Invoice Ever Arrives", "answer": "A local model running under Ollama produces no invoice, so its spend is invisible rather than zero. The useful number is what the same token volume would have cost on a hosted model, because that's what running it yourself is worth. Culpa, a local-first LLM cost, margin, and forecast ledger, counts local calls through the same route as any provider and prices that comparison.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$44.88", "claim": "modelled monthly hosted-equivalent value of a local Llama 3.3 70B deployment", "confidence": "calculated", "source": "(60 x $0.59) + (12 x $0.79) per million, using real Groq Llama 3.3 70B rates from the price book, effective 2026-07-02 and re-verified 2026-08-01. Token volumes are modelled." }, { "value": "$0.00", "claim": "provider invoice for local inference, which is why the counterfactual is the only comparable number", "confidence": "calculated", "source": "No provider bills a model running on your own hardware, so the invoice line is zero by construction." } ], "sources": [ { "label": "Ollama OpenAI compatibility", "url": "https://docs.ollama.com/api/openai-compatibility" }, { "label": "Culpa docs: connect your providers", "url": "https://app.getculpa.com/docs/how-to/connect-your-providers" }, { "label": "Groq pricing", "url": "https://groq.com/pricing" } ] }, { "slug": "/track-llm-cost-openai", "url": "https://getculpa.com/track-llm-cost-openai", "textUrl": "https://getculpa.com/track-llm-cost-openai/llms.txt", "pageType": "integration", "intent": "HOW_TO", "primaryKeyword": "track llm cost openai", "title": "Track LLM Cost on OpenAI, and the Streaming Usage Trap", "answer": "A streamed OpenAI response carries token usage only when the request asks for it, so a streaming app that leaves that setting off records calls with no numbers attached. Culpa, a local-first LLM cost, margin, and forecast ledger, counts what arrives and marks a call with no usage as a floor rather than guessing at it, so the gap shows up as a flag instead of as a quiet zero.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$45,000", "claim": "modelled monthly spend that goes unpriced when streamed calls carry no usage event", "confidence": "calculated", "source": "3,000,000 calls x $0.015, using real GPT-5.4 rates per 1k from the price book, effective 2026-07-02. Call volume and token counts are modelled." }, { "value": "$0.00", "claim": "what the ledger records for a call with no usage, alongside an incomplete flag", "confidence": "calculated", "source": "An absent usage field is stored as missing rather than guessed, so the priced total for that call is zero and the completeness flag carries the meaning." } ], "sources": [ { "label": "Culpa docs: connect your providers", "url": "https://app.getculpa.com/docs/how-to/connect-your-providers" }, { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/track-llm-cost-openrouter", "url": "https://getculpa.com/track-llm-cost-openrouter", "textUrl": "https://getculpa.com/track-llm-cost-openrouter/llms.txt", "pageType": "integration", "intent": "HOW_TO", "primaryKeyword": "track llm cost openrouter", "title": "Track LLM Cost on OpenRouter: Price What Answered", "answer": "A routing layer can answer with a different model than the one you requested, so pricing the request rather than the response produces a number that matches nothing. Culpa, a local-first LLM cost, margin, and forecast ledger, reads the model the response says served the call and prices that one, which is the only version of the arithmetic your invoice agrees with.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$30,800 to $35,000", "claim": "modelled monthly routed spend priced on the served model against priced on the requested one", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at real Claude Opus 4.7 and Sonnet 4.6 rates from the price book, effective 2026-07-02. A range because call volume and the fallback share are modelled." }, { "value": "$4,200", "claim": "modelled overstatement from keying a month of routed cost on the requested model", "confidence": "calculated", "source": "$35,000 minus $30,800, from the teardown arithmetic at real price-book rates." } ], "sources": [ { "label": "Culpa docs: connect your providers", "url": "https://app.getculpa.com/docs/how-to/connect-your-providers" }, { "label": "OpenRouter models", "url": "https://openrouter.ai/models" }, { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/track-llm-cost-python", "url": "https://getculpa.com/track-llm-cost-python", "textUrl": "https://getculpa.com/track-llm-cost-python/llms.txt", "pageType": "integration", "intent": "HOW_TO", "primaryKeyword": "track llm cost python", "title": "Track LLM Cost in Python Without Sending Your API Key", "answer": "Culpa's Python SDK reports the raw request and response to a Culpa server on your own machine, which prices them exactly as it prices a call captured on the request path. Your provider key never travels, because the SDK sends a local fingerprint instead. Culpa, a local-first LLM cost, margin, and forecast ledger, meters every reported call without adding a single token to your bill.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$0.00", "claim": "provider tokens added by capture itself, at any traffic volume", "confidence": "calculated", "source": "The SDK reports an already-completed call and makes no model call of its own, so the token count it adds is zero." }, { "value": "$22,500", "claim": "modelled monthly provider spend the capture layer measures without adding to", "confidence": "calculated", "source": "1,500,000 calls x $0.015, using real GPT-5.4 rates per 1k from the price book, effective 2026-07-02. Call volume and token counts are modelled." } ], "sources": [ { "label": "Culpa docs: tag your calls", "url": "https://app.getculpa.com/docs/how-to/tag-your-calls" }, { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/track-llm-cost-typescript", "url": "https://getculpa.com/track-llm-cost-typescript", "textUrl": "https://getculpa.com/track-llm-cost-typescript/llms.txt", "pageType": "integration", "intent": "HOW_TO", "primaryKeyword": "track llm cost typescript", "title": "Track LLM Cost in TypeScript When Your App Runs in the Cloud", "answer": "Culpa's TypeScript SDK reports calls to a Culpa server, and the default install binds to localhost only. A cloud-hosted app therefore has no route home, so its spend goes unattributed until a relay gives it a mailbox. Culpa, a local-first LLM cost, margin, and forecast ledger, traces every call it receives, and the question worth asking first is which of your calls can reach it.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$140 to $7,140", "claim": "modelled monthly spend a local-only ledger captures against the full bill", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at real Claude Haiku 4.5 rates from the price book, effective 2026-07-02. A range because call volumes are modelled." }, { "value": "2%", "claim": "modelled capture coverage for a cloud-hosted product with no relay armed", "confidence": "calculated", "source": "$140 divided by $7,140, from the teardown arithmetic." } ], "sources": [ { "label": "Culpa docs: deploy a relay", "url": "https://app.getculpa.com/docs/how-to/deploy-a-relay" }, { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/track-llm-cost-vercel-ai-sdk", "url": "https://getculpa.com/track-llm-cost-vercel-ai-sdk", "textUrl": "https://getculpa.com/track-llm-cost-vercel-ai-sdk/llms.txt", "pageType": "integration", "intent": "HOW_TO", "primaryKeyword": "track llm cost vercel ai sdk", "title": "Track LLM Cost in the Vercel AI SDK, and the Streaming Gap", "answer": "Culpa's Vercel AI SDK adapter is middleware you pass to wrapLanguageModel, and it reports token usage from the SDK's own usage object. It wraps generate calls, and streaming capture stays out of scope for now, so a streaming chat product records only part of its spend. Culpa, a local-first LLM cost, margin, and forecast ledger, meters what the adapter reports and shows the gap against your invoice.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$5,400 to $30,000", "claim": "modelled monthly spend captured through generate-only middleware against the real invoice", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at real GPT-5.4 rates per 1k from the price book, effective 2026-07-02. A range because call volume and the streaming share are modelled." }, { "value": "18%", "claim": "modelled capture coverage on a product where 82% of responses stream", "confidence": "calculated", "source": "360,000 captured calls divided by 2,000,000 total, from the teardown arithmetic." } ], "sources": [ { "label": "Vercel AI SDK: language model middleware", "url": "https://ai-sdk.dev/docs/ai-sdk-core/middleware" }, { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/unattributed-llm-spend", "url": "https://getculpa.com/unattributed-llm-spend", "textUrl": "https://getculpa.com/unattributed-llm-spend/llms.txt", "pageType": "problem", "intent": "AUDIT", "primaryKeyword": "unattributed llm spend", "title": "Unattributed LLM Spend: The Bucket That Quietly Grows", "answer": "Unattributed spend is every dollar your breakdown can't assign to a feature, user or customer. It grows because new call sites ship without tags, and it hides your worst waste. Culpa, a local-first LLM cost, margin, and forecast ledger, tracks the untagged share as its own number, so the gap in your attribution stays visible.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "30% to 35%", "claim": "modelled share of a monthly invoice sitting outside the feature breakdown", "confidence": "estimated", "source": "$3,000 unattributed against a $9,000 modelled invoice, from the teardown arithmetic. Published as a scenario because no customer data backs the split." }, { "value": "$1,900 to $3,000", "claim": "modelled unattributed spend, background jobs alone versus the full gap", "confidence": "estimated", "source": "Both endpoints from the modelled breakdown above. A range because it brackets one known contributor against the total." } ], "sources": [ { "label": "Groq pricing", "url": "https://groq.com/pricing" } ] }, { "slug": "/what-is-a-deep-trace", "url": "https://getculpa.com/what-is-a-deep-trace", "textUrl": "https://getculpa.com/what-is-a-deep-trace/llms.txt", "pageType": "metric", "intent": "LEARN", "primaryKeyword": "deep trace", "title": "What Is a Deep Trace? One Conversation, Down to the Prompt", "answer": "A deep trace opens one conversation all the way down to the prompt, so you can read what actually happened rather than what a chart summarised. Culpa, a local-first LLM cost, margin, and forecast ledger, meters these separately from ordinary tracking because they're the expensive question, and an entry plan carries a few hundred a month.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "250 to 1,200", "claim": "deep traces carried per month across the three published tiers", "confidence": "estimated", "source": "Culpa's published plan table, read 2026-08-02, and the same figures in the shipped plan configuration. A range because it spans the entry and top tiers." }, { "value": "0.14%", "claim": "modelled share of a month's conversations an entry allowance can open", "confidence": "calculated", "source": "250 deep traces divided by 180,000 modelled conversations, from the teardown arithmetic." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/what-is-a-tracked-call", "url": "https://getculpa.com/what-is-a-tracked-call", "textUrl": "https://getculpa.com/what-is-a-tracked-call/llms.txt", "pageType": "metric", "intent": "LEARN", "primaryKeyword": "tracked call", "title": "What Is a Tracked Call? The Unit a Metered Plan Counts", "answer": "A tracked call is one model request a cost ledger recorded, and it isn't a flat unit. Culpa, a local-first LLM cost, margin, and forecast ledger, counts one credit for the call plus one more for each complete 10,000 tokens it moved, so a large call costs several times a small one. The meter counts work rather than requests.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "1 to 5 credits", "claim": "modelled cost of a single tracked call across a 3,000 to 45,000 token range", "confidence": "estimated", "source": "One credit per call plus one per complete 10,000 combined tokens, applied to the three modelled token profiles in the teardown. A range because the profiles are modelled." }, { "value": "1,000,000", "claim": "credits the entry tier carries each month, as published", "confidence": "calculated", "source": "Culpa's published plan table, read 2026-08-02, and the same figure in the shipped plan configuration." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/what-is-blended-rate", "url": "https://getculpa.com/what-is-blended-rate", "textUrl": "https://getculpa.com/what-is-blended-rate/llms.txt", "pageType": "metric", "intent": "LEARN", "primaryKeyword": "blended rate", "title": "What Is a Blended Rate? Averaging LLM Cost Across Models", "answer": "A blended rate is your average cost per thousand tokens across every model you run, weighted by how much traffic each one takes. It's one number for a mixed fleet. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each call at its own model rate first and blends afterwards, so the average never hides which model moved.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$0.0011 to $0.0015 per 1k", "claim": "modelled blended input rate before and after a mix shift at identical token volume", "confidence": "calculated", "source": "Weighted averages of real Claude Haiku 4.5 and Claude Sonnet 5 input rates from the price book, effective 2026-07-02. Token splits are modelled, arithmetic shown in the teardown." }, { "value": "36%", "claim": "bill increase caused by mix shift alone, with volume and list prices unchanged", "confidence": "calculated", "source": "($15 - $11) / $11, from the two monthly totals in the teardown." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/what-is-cost-attribution", "url": "https://getculpa.com/what-is-cost-attribution", "textUrl": "https://getculpa.com/what-is-cost-attribution/llms.txt", "pageType": "metric", "intent": "LEARN", "primaryKeyword": "cost attribution", "title": "What Is LLM Cost Attribution? Linking Spend to Its Cause", "answer": "LLM cost attribution links every dollar of model spend to the call, conversation, feature, and customer that caused it. It's a capture-time decision, because the identifiers can't be recovered afterwards. Culpa, a local-first LLM cost, margin, and forecast ledger, attaches those identifiers as calls leave and maps the chain from dollar through to customer.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$300 to $380", "claim": "modelled monthly spend increase traced through four attribution levels", "confidence": "estimated", "source": "Modelled scenario used to contrast grouping levels, priced against real GPT-5.4 rates from the price book, effective 2026-07-02. Published as a range because no customer data backs the point figure." }, { "value": "85% to 95%", "claim": "share of the modelled increase attributable to a single product surface", "confidence": "estimated", "source": "Modelled distribution across surfaces in the teardown above. A range, because it illustrates the shape rather than a measured result." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/what-is-cost-cross-check", "url": "https://getculpa.com/what-is-cost-cross-check", "textUrl": "https://getculpa.com/what-is-cost-cross-check/llms.txt", "pageType": "metric", "intent": "LEARN", "primaryKeyword": "cost cross check", "title": "What Is a Cost Cross-Check? Calculated Against Reported", "answer": "A cost cross-check compares the figure you calculated from tokens and rates against the figure the provider reported for the same call, and flags a difference bigger than display rounding explains. Culpa, a local-first LLM cost, margin, and forecast ledger, reconciles the two per model, so a wrong rate surfaces as a disagreement rather than as a plausible total.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$0.000002", "claim": "per-call tolerance that separates provider display rounding from a genuine pricing error", "confidence": "calculated", "source": "An absolute per-call tolerance chosen to sit just above provider-side display rounding, applied to the absolute difference between calculated and reported cost." }, { "value": "$310.00", "claim": "modelled monthly disagreement on one model, from a rate entered 10% low", "confidence": "calculated", "source": "$3,410.00 reported minus $3,100.00 calculated, from the teardown arithmetic. The call volumes are modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/what-is-cost-floor", "url": "https://getculpa.com/what-is-cost-floor", "textUrl": "https://getculpa.com/what-is-cost-floor/llms.txt", "pageType": "metric", "intent": "LEARN", "primaryKeyword": "cost floor", "title": "What Is a Cost Floor? A Number That Admits What It Missed", "answer": "A cost floor is the lowest defensible figure for a call, recorded when something needed to price it fully never arrived, and flagged as incomplete rather than estimated into looking finished. Culpa, a local-first LLM cost, margin, and forecast ledger, counts a floor as a floor, so a total you can act on stays separate from a total that quietly filled its own gaps.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$63,000", "claim": "modelled monthly spend from fully priced calls, excluding every floored row", "confidence": "calculated", "source": "4,200,000 calls x $0.015, using real GPT-5.4 rates per 1k from the price book, effective 2026-07-02. Call volume and the mix of gaps are modelled." }, { "value": "16%", "claim": "modelled share of calls contributing nothing to a total that reads as complete", "confidence": "calculated", "source": "800,000 floored rows divided by 5,000,000 calls, from the teardown arithmetic." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/what-is-cost-per-agent", "url": "https://getculpa.com/what-is-cost-per-agent", "textUrl": "https://getculpa.com/what-is-cost-per-agent/llms.txt", "pageType": "metric", "intent": "LEARN", "primaryKeyword": "cost per agent", "title": "What Is Cost per Agent? The Whole Run, Not the First Call", "answer": "Cost per agent is the total model spend for one complete autonomous run, counting every sub-call, tool step and retry the agent made before it answered. One run isn't one call, it's a tree. Culpa, a local-first LLM cost, margin, and forecast ledger, traces each call back to the run that spawned it and prices the whole tree as a single unit.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$1.19", "claim": "modelled cost of one complete agent run across 18 calls on Claude Opus 4.8", "confidence": "calculated", "source": "Sum of the teardown arithmetic at real Claude Opus 4.8 rates per 1k tokens from the price book, effective 2026-07-02. The call tree and token volumes are modelled." }, { "value": "3%", "claim": "share of a modelled run's total cost carried by the opening planner call", "confidence": "calculated", "source": "$0.040 divided by $1.19, from the teardown arithmetic." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/what-is-cost-per-agent-step", "url": "https://getculpa.com/what-is-cost-per-agent-step", "textUrl": "https://getculpa.com/what-is-cost-per-agent-step/llms.txt", "pageType": "metric", "intent": "LEARN", "primaryKeyword": "cost per agent step", "title": "What Is Cost per Agent Step? Why Step 20 Costs More", "answer": "Cost per agent step is the model spend for one call inside an agent run, and it isn't constant across the run. Each step carries the transcript of the steps before it, so input grows as the agent works. Culpa, a local-first LLM cost, margin, and forecast ledger, prices every step individually and shows where in a run the money actually goes.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$0.0035 to $0.0206", "claim": "modelled cost of a single agent step, first step against twentieth, on Claude Haiku 4.5", "confidence": "estimated", "source": "Both endpoints at real Claude Haiku 4.5 rates per 1k tokens from the price book, effective 2026-07-02. A range because the per-step token volumes are modelled." }, { "value": "$0.241", "claim": "modelled total cost of a 20-step run before any step-budget increase", "confidence": "calculated", "source": "211k input tokens x $0.001 plus 6k output tokens x $0.005 per 1k, from the teardown arithmetic." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/what-is-cost-per-conversation", "url": "https://getculpa.com/what-is-cost-per-conversation", "textUrl": "https://getculpa.com/what-is-cost-per-conversation/llms.txt", "pageType": "metric", "intent": "LEARN", "primaryKeyword": "cost per conversation", "title": "What Is Cost per Conversation? Definition and Worked Math", "answer": "Cost per conversation is the total model spend for one complete exchange, every call, retry, and tool step included. It's a sum, not an average. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each call from a versioned price book and rolls it up to the conversation that spent it, so you can name the exchange behind a spike.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$0.0416", "claim": "cost of a five-turn support exchange with history resent each turn", "confidence": "calculated", "source": "Sum of five turns at Claude Sonnet 5 introductory rates ($0.002 per 1k input, $0.010 per 1k output, price book effective 2026-07-02). Full arithmetic shown in the teardown." }, { "value": "4.4x", "claim": "growth in input tokens between turn one and turn five of that exchange", "confidence": "calculated", "source": "4,400 input tokens at turn five divided by 1,000 at turn one, from the modelled token profile above." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/what-is-cost-per-customer", "url": "https://getculpa.com/what-is-cost-per-customer", "textUrl": "https://getculpa.com/what-is-cost-per-customer/llms.txt", "pageType": "metric", "intent": "LEARN", "primaryKeyword": "cost per customer", "title": "What Is Cost per Customer? Why It Isn't Cost per User", "answer": "Cost per customer is model spend attributed to a paying account rather than to one person inside it. In B2B a single customer can hold dozens of users, so cost per user hides the number your contract actually has to cover. Culpa, a local-first LLM cost, margin, and forecast ledger, tags each call to an account and rolls seat activity up to the customer paying the invoice.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$2.85", "claim": "modelled monthly model cost for one active seat on GPT-5.4 mini", "confidence": "calculated", "source": "(2 x $0.75) + (0.3 x $4.50) per million, using real GPT-5.4 mini rates from the price book, effective 2026-07-02. Per-seat token volumes are modelled." }, { "value": "$11.40 to $131.10", "claim": "modelled monthly cost of one account as it grew from 4 active seats to 46 on an unchanged contract", "confidence": "estimated", "source": "Both endpoints from the teardown arithmetic at real GPT-5.4 mini rates. A range because the seat expansion curve is modelled." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/what-is-cost-per-feature", "url": "https://getculpa.com/what-is-cost-per-feature", "textUrl": "https://getculpa.com/what-is-cost-per-feature/llms.txt", "pageType": "metric", "intent": "LEARN", "primaryKeyword": "cost per feature", "title": "What Is Cost per Feature? Tagging LLM Spend to Product Surfaces", "answer": "Cost per feature is model spend grouped by the product surface that caused it, rather than by model or by day. It answers which part of your product burns the budget. Culpa, a local-first LLM cost, margin, and forecast ledger, tags every call to a feature at request time and rolls the spend up, so you can rank surfaces by what they actually cost.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$169", "claim": "modelled monthly spend across three surfaces sharing two models", "confidence": "calculated", "source": "Sum of three feature totals using real GPT-5.4 and GPT-5.4 mini rates from the price book (effective 2026-07-02). Token volumes are modelled, arithmetic shown in the teardown." }, { "value": "$50", "claim": "modelled monthly cost of a low-traffic report-drafting surface", "confidence": "calculated", "source": "(2,000 x $0.0025) + (3,000 x $0.015), GPT-5.4 rates per 1k tokens from the price book." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/what-is-cost-per-prompt-version", "url": "https://getculpa.com/what-is-cost-per-prompt-version", "textUrl": "https://getculpa.com/what-is-cost-per-prompt-version/llms.txt", "pageType": "metric", "intent": "LEARN", "primaryKeyword": "cost per prompt version", "title": "What Is Cost per Prompt Version? Pricing Every Edit", "answer": "Cost per prompt version is model spend split by which revision of a prompt served each call. Without a version stamped on the call at request time, an edit and the cost change it caused can't be connected afterwards. Culpa, a local-first LLM cost, margin, and forecast ledger, names the prompt version behind every call, so an edit reads as a priced change rather than a mystery.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$0.00465 to $0.009075", "claim": "modelled cost per call across two versions of one prompt on Gemini 2.5 Pro", "confidence": "estimated", "source": "Both endpoints at real Gemini 2.5 Pro rates per 1k tokens from the price book, effective 2026-07-02. A range because the token counts on each version are modelled." }, { "value": "$3,983", "claim": "modelled monthly increase produced by a single prompt edit at 900,000 calls", "confidence": "calculated", "source": "(900,000 x $0.009075) minus (900,000 x $0.00465), using real Gemini 2.5 Pro rates per 1k tokens. Call volume and token counts are modelled." } ], "sources": [ { "label": "Gemini API pricing", "url": "https://ai.google.dev/gemini-api/docs/pricing" } ] }, { "slug": "/what-is-cost-per-user", "url": "https://getculpa.com/what-is-cost-per-user", "textUrl": "https://getculpa.com/what-is-cost-per-user/llms.txt", "pageType": "metric", "intent": "LEARN", "primaryKeyword": "cost per user", "title": "What Is Cost per User for an AI Product? Median vs P90", "answer": "Cost per user is model spend attributed to individual users, read as a distribution rather than an average. On AI products the top decile routinely costs many times the median. Culpa, a local-first LLM cost, margin, and forecast ledger, tags each call to a user and reports the spread, so your pricing floor comes from the heavy tail rather than the mean.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$0.51 mean, $0.135 median", "claim": "modelled cost per active user across a long-tailed distribution of 1,000 users", "confidence": "calculated", "source": "Totals computed at real Gemini 2.5 Flash rates per 1k tokens from the price book, effective 2026-07-02. The user distribution is modelled, arithmetic shown in the teardown." }, { "value": "53%", "claim": "share of modelled total spend generated by the heaviest 1% of users", "confidence": "calculated", "source": "$270 from ten heavy users divided by the $513 modelled total, from the teardown arithmetic." } ], "sources": [ { "label": "Gemini API pricing", "url": "https://ai.google.dev/gemini-api/docs/pricing" } ] }, { "slug": "/what-is-effective-cost", "url": "https://getculpa.com/what-is-effective-cost", "textUrl": "https://getculpa.com/what-is-effective-cost/llms.txt", "pageType": "metric", "intent": "LEARN", "primaryKeyword": "effective cost", "title": "What Is Effective Cost? The Rate You Actually Pay per Token", "answer": "Effective cost is what you really pay per useful token once retries, failed calls, and cache behaviour are counted, rather than the list rate on a pricing page. It's almost always higher. Culpa, a local-first LLM cost, margin, and forecast ledger, meters every call including the ones that failed, so your effective cost reflects spend rather than intent.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$0.0031 per 1k", "claim": "modelled effective input rate against a $0.0025 list rate", "confidence": "calculated", "source": "$31.00 total billed divided by 10,000 thousand-token units of useful input. List rate from the GPT-5.4 price book row, effective 2026-07-02. Retry and discard percentages are modelled." }, { "value": "24%", "claim": "gap between list rate and effective rate in the modelled month", "confidence": "calculated", "source": "($0.0031 - $0.0025) / $0.0025, from the arithmetic in the teardown." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/what-is-forecast-accuracy", "url": "https://getculpa.com/what-is-forecast-accuracy", "textUrl": "https://getculpa.com/what-is-forecast-accuracy/llms.txt", "pageType": "metric", "intent": "LEARN", "primaryKeyword": "forecast accuracy", "title": "What Is Forecast Accuracy? Two Numbers, Never One", "answer": "Forecast accuracy is two measurements taken together: how far the middle estimate sat from what actually happened, and whether reality landed inside the upper band. Either one on its own can flatter a useless forecast. Culpa, a local-first LLM cost, margin, and forecast ledger, forecasts every period and grades it once the period closes, so the record is scored rather than remembered.", "lastReviewed": "2026-08-02", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "3.23% to 31.03%", "claim": "modelled median forecast error across four graded periods", "confidence": "estimated", "source": "Each endpoint is the absolute difference between the modelled median and actual, over the actual, from the teardown arithmetic. A range because the four periods are modelled." }, { "value": "2.3x", "claim": "modelled band width as a multiple of the median in the period that scored best on error", "confidence": "calculated", "source": "$700 divided by $300, from the month 4 figures in the teardown." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/what-is-gross-margin-per-customer", "url": "https://getculpa.com/what-is-gross-margin-per-customer", "textUrl": "https://getculpa.com/what-is-gross-margin-per-customer/llms.txt", "pageType": "metric", "intent": "LEARN", "primaryKeyword": "gross margin per customer", "title": "What Is Gross Margin per Customer for AI Products?", "answer": "Gross margin per customer is what an account pays you minus what serving it costs in model spend, per account rather than averaged. On usage-heavy AI products the spread between accounts is enormous. Culpa, a local-first LLM cost, margin, and forecast ledger, tags spend to a customer and reconciles it against revenue you type in or import.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$260", "claim": "modelled monthly model spend for a heavy account paying $200", "confidence": "calculated", "source": "(60,000 x $0.002) + (14,000 x $0.010) using real Claude Sonnet 5 introductory rates per 1k tokens from the price book, effective 2026-07-02. Usage volumes are modelled." }, { "value": "-30% to 96%", "claim": "modelled gross margin spread between two accounts on an identical plan", "confidence": "estimated", "source": "Range across the two modelled usage profiles above, both priced at real Claude Sonnet 5 rates. Shown as a range because the endpoints depend entirely on your own usage distribution." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/what-is-llm-burn-forecast", "url": "https://getculpa.com/what-is-llm-burn-forecast", "textUrl": "https://getculpa.com/what-is-llm-burn-forecast/llms.txt", "pageType": "metric", "intent": "LEARN", "primaryKeyword": "llm burn forecast", "title": "What Is an LLM Burn Forecast? Predicting Next Month's Bill", "answer": "An LLM burn forecast projects your model spend forward from your own measured usage rather than from a rate card. A credible one carries a range and states its assumptions. Culpa, a local-first LLM cost, margin, and forecast ledger, forecasts deterministically from your ledger and stores every projection so its accuracy can be scored later.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$700 to $950", "claim": "modelled forecast range for a launch month at 2,000 active users", "confidence": "estimated", "source": "Central figure $800, from driver-level arithmetic at real GPT-5.4 rates (price book, effective 2026-07-02) with a 24% effective-rate uplift. Shown as a range because launch-week traffic is the dominant uncertainty." }, { "value": "$0.22", "claim": "modelled steady-state model cost per active user per month", "confidence": "calculated", "source": "(40 x $0.0025) + (8 x $0.015), GPT-5.4 rates per 1k tokens from the price book. Per-user token volumes are modelled." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] }, { "slug": "/what-is-local-first-llm-observability", "url": "https://getculpa.com/what-is-local-first-llm-observability", "textUrl": "https://getculpa.com/what-is-local-first-llm-observability/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "local first llm observability", "title": "Local-First LLM Observability: Table Stakes, Not a Differentiator", "answer": "Local-first LLM observability means the cost ledger and its history live on infrastructure you control rather than a vendor's. Culpa, a local-first LLM cost, margin, and forecast ledger, runs that way by default. It's worth saying plainly that this is close to table stakes in the category rather than the reason to pick any one tool.", "lastReviewed": "2026-08-03", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "about 12 GB", "claim": "modelled size of twelve months of call-level cost history at 5 million calls a month", "confidence": "calculated", "source": "5,000,000 calls x 200 bytes per priced row = about 1 GB a month, so roughly 12 GB for a year. Row size and call volume are modelled. Compared against New Relic's published 100 GB free monthly ingest and Langfuse's published 90-day Core and 3-year Pro retention tiers, both read 2026-08-03, to show the binding constraint is retention policy rather than storage cost." } ], "sources": [ { "label": "New Relic pricing", "url": "https://newrelic.com/pricing" }, { "label": "Langfuse pricing", "url": "https://langfuse.com/pricing" } ] }, { "slug": "/what-is-prompt-cache-hit-rate", "url": "https://getculpa.com/what-is-prompt-cache-hit-rate", "textUrl": "https://getculpa.com/what-is-prompt-cache-hit-rate/llms.txt", "pageType": "metric", "intent": "LEARN", "primaryKeyword": "prompt cache hit rate", "title": "What Is Prompt Cache Hit Rate? And When Caching Costs More", "answer": "Prompt cache hit rate is the share of your input tokens that billed at the cached rate rather than the standard one. On some providers a cache write bills above the input rate, so a badly tuned cache costs more than none. Culpa, a local-first LLM cost, margin, and forecast ledger, meters cached, uncached, and cache-write tokens separately per call.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "1 read", "claim": "break-even point before a five-minute cache write pays for itself on Claude Haiku 4.5", "confidence": "calculated", "source": "Incremental write cost $0.001 divided by the $0.0036 per-read saving = 0.28, so the first read clears it. Claude Haiku 4.5 rates from the price book (effective 2026-07-02) with the published 1.25x write multiplier. Matches Anthropic's own statement that a five-minute cache pays off after one read." }, { "value": "25%", "claim": "cost penalty of rebuilding the cache on every call versus not caching at all", "confidence": "calculated", "source": "($0.005 - $0.004) / $0.004, from the per-call arithmetic shown in the teardown." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/what-is-token-accounting", "url": "https://getculpa.com/what-is-token-accounting", "textUrl": "https://getculpa.com/what-is-token-accounting/llms.txt", "pageType": "metric", "intent": "LEARN", "primaryKeyword": "token accounting", "title": "What Is Token Accounting? Counting LLM Spend Without Floats", "answer": "Token accounting is recording input, output, and cached tokens for every call, then pricing each category at its own rate in exact decimal arithmetic. The three categories bill differently, so one combined count can't produce a correct figure. Culpa, a local-first LLM cost, margin, and forecast ledger, meters all three per call and prices them from a versioned book.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$0.0037", "claim": "correct cost of a single cached Claude Haiku 4.5 call, priced by category", "confidence": "calculated", "source": "(1 x $0.001) + (2 x $0.0001) + (0.5 x $0.005), using Claude Haiku 4.5 rates per 1k tokens from the price book, effective 2026-07-02." }, { "value": "5%", "claim": "understatement produced by collapsing the three token categories into one count", "confidence": "calculated", "source": "($0.0037 - $0.0035) / $0.0037, from the two calculations shown in the teardown." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/what-to-cut-first", "url": "https://getculpa.com/what-to-cut-first", "textUrl": "https://getculpa.com/what-to-cut-first/llms.txt", "pageType": "problem", "intent": "LEARN", "primaryKeyword": "what to cut first llm cost", "title": "What to Cut First: Model Choice Beats Caching by 50 Points", "answer": "The levers that reduce model spend differ by an order of magnitude, and the biggest one is which model you run. Caching, trimming and retries matter and come second. Culpa, a local-first LLM cost, margin, and forecast ledger, prices your real traffic so the ordering is measured on your workload rather than assumed.", "lastReviewed": "2026-08-05", "effectiveDate": "2026-07-02", "calculatorMode": null, "costFigures": [ { "value": "80% against 30%", "claim": "modelled saving from changing model against caching most of the prompt, on the same month", "confidence": "calculated", "source": "A modelled million calls at 2,000 input and 500 output tokens on Claude Opus 5 at real rates of $5.00 and $25.00 per million costs $0.0225 a call, so $22,500.00. On Claude Haiku 4.5 at $1.00 and $5.00 the same shape costs $0.0045, so $4,500.00, an $18,000.00 or 80% saving. Caching 1,500 of the 2,000 input tokens on Opus 5 at the $0.50 per million cached rate gives $0.01575, so $15,750.00, a $6,750.00 or 30% saving. Rates are published, the volume, token shape and cacheable share are modelled." } ], "sources": [ { "label": "Anthropic pricing", "url": "https://platform.claude.com/docs/en/docs/about-claude/pricing" } ] }, { "slug": "/why-did-my-openai-bill-go-up", "url": "https://getculpa.com/why-did-my-openai-bill-go-up", "textUrl": "https://getculpa.com/why-did-my-openai-bill-go-up/llms.txt", "pageType": "problem", "intent": "AUDIT", "primaryKeyword": "why did my openai bill go up", "title": "Why Did My OpenAI Bill Go Up? The Four Causes, Ranked", "answer": "Your OpenAI bill went up for one of four reasons: more calls, longer calls, a shift toward a pricier model, or retries you paid for twice. Culpa, a local-first LLM cost, margin, and forecast ledger, prices every call from a versioned price book and names the conversation, feature, and customer behind the increase.", "lastReviewed": "2026-08-01", "effectiveDate": null, "calculatorMode": null, "costFigures": [ { "value": "$400 to $600", "claim": "modelled monthly spend before and after, priced at real GPT-5.4 rates", "confidence": "estimated", "source": "Baseline (40,000 x $0.0025) + (20,000 x $0.015) using GPT-5.4 rates per 1k tokens from the price book, effective 2026-07-02. Published as a range because the token volumes are modelled." }, { "value": "$0.0025 / $0.015 per 1k", "claim": "GPT-5.4 input and output rates used throughout the arithmetic", "confidence": "calculated", "source": "Culpa price book row for openai/gpt-5.4, effective 2026-07-02, verified against OpenAI API pricing." } ], "sources": [ { "label": "OpenAI API pricing", "url": "https://developers.openai.com/api/docs/pricing" } ] } ] }