Guides / llm cost in new relic
How to see LLM cost in New Relic
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.
Why this happens
New Relic ships AI Monitoring, which it describes as scaling AI with visibility into performance and token usage, and which answers whether users are waiting too long, whether there's a recent spike in token usage, and whether feedback is clustering around certain topics. Token usage is there. What sits underneath sits the pricing model, and for LLM work that model matters more than usual. New Relic publishes 100 GB of free ingest a month and then $0.40 per GB on standard data, or $0.60 per GB on Data Plus, with an extra $0.05 per GB per month to store in the EU. LLM traces are unusually heavy because the interesting parts are prompts and responses, so the more useful your instrumentation becomes the more it costs, and that cost is charged by volume rather than by model spend. The result is two bills growing for different reasons, and only one of them is on your provider's invoice.
What this usually looks like
- Your ingest bill grew the month you added AI monitoring and your model bill didn't.
- Prompts and responses are the useful part of the trace and the expensive part of the bill.
- You sampled traces to control ingest and now cost attribution has holes in it.
- Token usage is visible and cost per customer isn't.
- Nobody priced data residency, which is a per-gigabyte line of its own.
Free, no card, no account
Run the free Cost Leak Scan
It shows your most expensive conversation before you install anything.
Mistakes that cost the most
| Mistake | Why it hurts | Do instead |
|---|---|---|
| Instrumenting full prompts and responses without pricing the ingest. | New Relic bills $0.40 per GB beyond 100 GB free, and prompt text is the heaviest thing in a trace. | Decide what has to be stored, and price the volume before turning it on. |
| Sampling traces to save ingest and keeping cost reporting. | Sampled traces give a sampled cost, so the total quietly becomes an estimate nobody labelled. | Keep cost metering complete even where tracing is sampled. They're different jobs on the same call. |
| Treating token usage as cost. | Tokens become money only through a rate that changes by model, tier and date. | Price tokens against a versioned rate book, so a past month recomputes as it was billed. |
| Choosing EU storage without costing it. | New Relic publishes an extra $0.05 per GB per month for it, on top of the per-GB ingest rate. | Treat residency as a priced line rather than a checkbox, because it multiplies with volume. |
Run this check tonight
- Find your monthly ingest in GB, and the month it changed.
- Work out what share of that volume is prompt and response text.
- Check whether your traces are sampled, and whether your cost figures inherit that sampling.
- Price 100 GB of overage at your plan's rate and compare it to your model spend.
What LLM traces cost to store
New Relic publishes 100 GB of free ingest a month, then $0.40 per GB on standard data and $0.60 on Data Plus, read from its pricing page on 2026-08-03. Take a product logging full prompts and responses at a modelled 8 KB per call across 5 million calls a month, and price the ingest beside the model spend on Claude Haiku 4.5 at $1.00 and $5.00 per million. Volumes are modelled and the rates are published.
The free allowance covers a lot until the traces get genuinely useful, and then the meter starts. That crossover is worth finding deliberately rather than discovering on an invoice.
Every number, with its confidence and source
| Figure | What it means | Confidence | Source |
|---|---|---|---|
| $120.00 per month | modelled New Relic ingest for 400 GB of LLM trace text on standard data pricing | calculated | 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. |
| $0.40 to $0.60 per GB | New Relic's published data ingest rates beyond the free allowance | provider-reported | 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. |
What a generic answer can’t know
New Relic answers whether the AI layer is behaving and how many tokens went through it, which is what AI Monitoring is for. Two things sit outside it. The first is pricing: token usage becomes money only through a rate that varies by model, by tier and by date, and a monitoring platform has no reason to carry a versioned rate book. The second is the join to revenue, which lives in a billing system and never reaches telemetry. Culpa carries both, and keeps the ledger on your own infrastructure, which also means metering every call is a disk decision rather than an ingest charge. That matters here more than elsewhere, because the platform you would otherwise meter with bills by the gigabyte.
Questions founders ask next
Does New Relic track LLM cost?
It ships AI Monitoring, which it describes as visibility into AI performance and token usage, and which can show a recent spike in token usage. Token usage is tracked. Turning tokens into money needs a rate book that varies by model, tier and date, which isn't what a monitoring platform carries.
What does New Relic charge for data ingest?
It publishes 100 GB free each month, then $0.40 per GB on standard data or $0.60 per GB on Data Plus, with an additional $0.05 per GB per month to store data in the EU. Read from its pricing page on 2026-08-03.
Why does that matter for LLM observability specifically?
Because the useful part of an LLM trace is text. Prompts, responses, retrieval context and tool payloads are far heavier than a typical HTTP span, so the instrumentation that makes tracing worth having is also what moves a per-gigabyte meter.
Can I sample traces and still get accurate costs?
Not from the traces. A sampled trace gives a sampled cost, and the total becomes an estimate that nobody labelled as one. Keep cost metering complete even where tracing is sampled, because they're different jobs performed on the same call.
On your infrastructure
Culpa runs on your infrastructure. Your prompts and responses never leave it. Culpa counts calls to run your plan, and it fails open, so if it ever breaks your app keeps running.
How Culpa works
Find the culprit. Not just the total.
Your dashboard shows what you spent. It stops short of who spent it. Culpa shows the conversation, the user and the feature behind it.
Your prompts stay local.
Culpa runs on your own infrastructure. What you send to a model reaches us at no point.
Every dollar has a name.
Follow any charge to the conversation, the user, the feature and the customer behind it.
See the bill before it lands.
Cost your next feature before you ship it. You get the likely bill and the worst case, at best, median, p90 and p99.
Three steps to your first answer.
Change one base URL.
Or drop in the Python or TypeScript library.
Find your most expensive conversation.
In the first session, not the first week.
Cost your next feature before you ship it.
Why the bill went up
Example dashboardCalls traced
418,209
across 3 projects
Spend this week
$378.41
+ $182 vs last week
Failed calls
312
74% retried, and you paid for all of them
+ $182 this week traced to one culprit
Spend over 14 days
Most expensive users
Next week forecast
Graded against reality. Accuracy shown as results land.
Keep reading
Sources: New Relic AI Monitoring, New Relic pricing. Last reviewed 2026-08-03, rates effective 2026-07-02. Plain text version.