Pricing
Every model is priced 20% below its provider’s list price. You pay in ETH on Robinhood Chain, request by request, for the tokens each one uses. There is no subscription.
Model prices
Reading the gateway’s price table…
Per million tokens, in US dollars. The struck-through figure is the provider’s own price.
Cost calculator
Reading the gateway’s price table…
The same integer arithmetic the gateway and evane-clientd use: each request is priced from its token counts and rounded up to the next micro-dollar and the next wei.
Optional. How long would it last?
How a request is charged
1
A request holds its maximum cost
While it runs, the gateway holds the most it could cost against your account: its input, estimated from its size, plus its output limit (max_tokens, or 4,096 when unset).
2
You pay for what it used
When the provider reports the tokens used, the gateway charges for those, never more than the hold. If the provider returns an error, the request costs nothing.
3
Your daemon checks every charge
evane-clientd recomputes each charge from the published prices and signs a receipt only for what it could verify.
4
Settled in batches on Robinhood Chain
Every 100 requests or 5 minutes, the gateway settles the lower of your latest receipt and what it charged. Its validator pays the gas.
The prices and the ETH rate are part of the gateway’s attested configuration: they cannot change without changing the compose hash your daemon pins, so a price change is always a new deployment you can see. The arithmetic is in the docs, and every settlement is on the vault explorer.