Sevren pricing.
Prices come from the live model catalog. Paid models are billed per token: a holder's free try comes first, then a holder plan's free tokens of the day on that model, then your account credit.
Model prices
Loading prices from the model catalog…
How billing works
Each model has two prices in USD per 1M tokens: one for input (the prompt, including earlier messages in the conversation) and one for output (the reply). A request costs its prompt tokens times the input price plus its output tokens times the output price, divided by 1,000,000.
Reasoning models, tagged reasoning in the table, think before they answer. Their thinking tokens are billed as output tokens, at the output price, and count toward the output limit; a short answer can therefore cost more than its visible text suggests. Reasoning effort is set to high by Sevren and cannot be changed per request. The usage of every response includes these tokens.
A request whose model has not started answering within 180 seconds is stopped with model_not_responding and costs nothing.
Credit is a USD balance on your account. While a request runs, the most it can cost is reserved; when it finishes, only the tokens it used are charged. Output is capped by what your free tokens and credit can pay for, and a request they cannot cover at all is refused. Your balance never goes negative.
Your session token (sk-sv-user-…) is also an API key for https://api.sevren.xyz/v1, billed the same way. It is valid for 14 days.
Holder plans and the free try
Holding the Sevren token gives your account free tokens on the paid models, in amounts set per model: a one-time free try and, with a plan, free tokens every day. Free tokens pay for tokens before your credit does.
Loading holder plans…
Each model has its own allowances, with prompt and output tokens counted separately.
Free tokens pay only for paid models, before any credit: the free try first, then the plan's free tokens of the day. When a request needs more than what is left on that model, free tokens cover part of it and credit pays the rest at the model's prices; other models keep their own allowances. Plan tokens reset every day at 00:00 UTC and unused ones do not carry over. The free try does not reset. Holder benefits has the details.
To get a plan or the free try, sign in with a wallet that holds the token. The free try is counted per wallet, once. The holding is measured as a share of the token's total supply and is re-checked regularly; you can also refresh it from the account page. The inference service only ever sees an account pseudonym.
Adding credit
You add credit on the account page with a direct payment of ETH or USDG to the Sevren payment contract on Robinhood Chain. After 900 block confirmations, usually 1 to 2 minutes later, its full value becomes account credit: USDG one to one, ETH at the Chainlink ETH/USD price at that moment. Payments explains the flow.
Payments are final: they cannot be withdrawn or refunded. A payment is one transaction, plus an approval first when USDG needs one; network fees are paid in ETH.
On-chain payments are public: anyone can see the paying wallet, the amount and the time, and Sevren links each payment to the account it credits by its payment reference. Sevren does not hide your IP address; see Privacy.