Guides / tencent hy3 preview pricing
Tencent HY3 Preview pricing on OpenRouter
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.
Why this happens
A preview rate is a marketing position as much as a price. It exists to get a model tried, and the vendor keeps the right to move it when the model leaves preview. Building unit economics on one is fine, and forgetting you did isn't. The failure is quiet: your cost per customer was computed against a rate that later changes, the model keeps working, and the margin you quoted to a board goes stale without a single error appearing anywhere.
What this usually looks like
- A cost-per-customer figure rests on a preview or beta rate nobody flagged as temporary.
- Nobody knows which models in your stack are on preview pricing today.
- A margin number was computed once and has never been re-priced.
- Your cheapest model is carrying a large share of traffic on the strength of a preview rate.
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 |
|---|---|---|
| Treating a preview rate as a settled one. | It's the rate most likely to move, and it usually moves up when the model graduates. | Tag preview models in your own records and re-price them on a schedule. |
| Routing high volume to a preview model without a fallback priced. | If the rate moves you have a migration to run under time pressure and no comparison ready. | Price the runner-up now, so a repricing is a decision rather than a scramble. |
| Quoting a margin built on a preview rate without saying so. | The number is right today and silently wrong later, which is worse than being wrong now. | Carry the rate's date and status with the margin figure, the way Culpa carries confidence. |
Run this check tonight
- List every model in your stack and mark which are preview, beta or general availability.
- Total the monthly spend sitting on preview rates. That figure is your exposure.
- Price the same volume against the nearest generally available model.
- The gap between the two is what a graduation would cost you.
- Re-read the router's API monthly, because a preview rate changes without an announcement you'll see.
What a preview rate is worth if it graduates
Illustrative example
500M input and 100M output tokens a month on Tencent HY3 Preview at its real OpenRouter rate of $0.063 per million input and $0.21 output, effective 2026-07-02, against DeepSeek V4 Flash at $0.14 and $0.28 as the nearest generally available routed alternative. Volumes are modelled.
Forty-five dollars a month of the bill rests on a rate the vendor never promised to keep. That's a fine bet as long as somebody wrote down that it's a bet.
Every number, with its confidence and source
| Figure | What it means | Confidence | Source |
|---|---|---|---|
| $0.063 per million | Tencent HY3 Preview input rate through OpenRouter | calculated | 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. |
| $52.50 to $98.00 | modelled monthly cost of one workload on the preview rate and on the nearest alternative | estimated | Both endpoints from the teardown arithmetic at real rates. A range because the token volumes are modelled. |
What a generic answer can’t know
A rate card tells you today's price. It can't tell you what share of your own traffic depends on it, which is the number that decides whether a repricing is an inconvenience or a problem. That needs your calls joined to your models. Culpa measures it on your infrastructure, keeps your prompts and responses there, and counts the calls to run your plan.
Questions founders ask next
How much does Tencent HY3 Preview cost?
$0.063 per million input tokens and $0.21 output through OpenRouter, with cached input at $0.021. Read from OpenRouter's models API on 2026-08-03 and unchanged from the rate Culpa's book has held since 2026-07-02.
What happens to the rate when it leaves preview?
Nobody outside Tencent knows, and that's the point. A preview rate carries no commitment, so the useful move is to know what share of your spend depends on it and what the nearest alternative would cost.
Why is the cached discount smaller than on Claude or GPT?
Cached input here is $0.021 against $0.063, roughly a third rather than a tenth. The discount is set per model, and the cheap routed models in the book cluster around a third while the frontier models sit at a tenth.
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.
Free, no card, no account
Run the free Cost Leak Scan
It shows your most expensive conversation before you install anything.
Keep reading
Sources: OpenRouter models, Anthropic pricing. Last reviewed 2026-08-03, rates effective 2026-07-02. Plain text version.