Chutes runs open models as a pay-per-token inference API on GPUs supplied by miners. Its public API publishes, for every model ("chute"), the dollar value of each day's calls. This page multiplies that daily value by 365 to give an annualized run-rate, and sets it against the network's daily token and request volume.
What the dollar figure is. The usd_amount field of
api.chutes.ai/invocations/usage. Chutes' own API specification
(openapi.json, read ) defines it as
"aggregated usage data, which is the amount of revenue we would be receiving if no usage was free." It is the value of
all metered LLM calls at Chutes' list prices.
What it is not. It is not cash collected: usage that was free, discounted or covered by a plan is counted at list price, and Chutes publishes no split. It is not Chutes' take or profit: miners are paid in alpha emissions, not from this figure. Its daily call counts equal the request totals of Chutes' LLM statistics endpoint exactly, so it covers LLM traffic only. Chutes' pricing page (read ) states "We do not offer a free tier at this time", which bears on the current gap between list value and cash but does not close it.
How far back it goes. The endpoint returns only the latest ~12 days and accepts no date parameters, so the series exists only where it was read: daily reads from , plus one archived copy of the endpoint from March 2025 (Wayback Machine, 23 Mar 2025). Token and request counts have their own full history since January 2025.
Each complete UTC day's list-price usage value multiplied by 365. A rate, not a forecast; gaps are days no read captured.
| Window | Usage value per day | Annualized | Tokens per day | Requests per day | $ per million tokens | $ per 1,000 requests |
|---|
Input plus output tokens processed by all LLM chutes each day.
Calls to an LLM chute each day.
| Chute (model) | Usage value | Share | Requests | $ per 1,000 requests |
|---|
| Week of | Days | Usage value | Annualized | Requests | Tokens | $ per million tokens | Source |
|---|