A price before you run.
Capacity-based rates. A request-level ceiling. The numbers below use a mock asset to explain the proposed flow.
Maximum hold
For 1,000 input tokens and an output limit of 500 tokens.
Read how billing works- Input rate: 0.20 MOCK / million tokens
- Output rate: 0.60 MOCK / million tokens
- Quote lifetime: 30 seconds
- Exact model and prompt identity
- Rates fixed on acceptance
- Capacity remains subject to availability
- Wallet-signed spending ceiling
- No reserve until acceptance
Measured charge
200 output tokens used. 0.000180 MOCK released.
See billing details- Pay for measured usage on success
- Unused reservation returns to available funds
- Failed pilot requests are unbilled
- Keep a signed billing receipt
Illustrative six-decimal mock asset fixture from the build brief. These are not production prices or verified settlement-token decimals. Discounts depend on available capacity and a measured cost floor; none are guaranteed.
What sets the actual rate?
The proposed pricing engine responds to occupied serving slots, subject to a measured cost floor. The production model, settlement asset, benchmarked rates, and paid service are still being validated.
Can I withdraw unused funds?
The design returns available, unreserved paid funds to the same wallet, token, and network. OFFPEAK controls the prepaid treasury and processes withdrawals. The preview does not accept deposits.