A lower rate.
Not a different request.
Bulk purchasing is how we lower the price. These are the requirements we set for the model, metering, context and settings you receive.
We’ll share the applicable model versions, rates, limits and supporting documentation before you send API traffic. The standards below are requirements, not an independent certification.
The model you choose.
Your requested model and version must not be silently replaced. If that model is unavailable, you should receive an explicit error rather than an undisclosed substitute.
No inflated token counts.
Usage categories and counts follow provider-reported metering. Charges use TokenAir’s published rates. Input, output, cache usage and any other billable categories are identified separately, where applicable.
No hidden cuts.
We require your submitted context to be forwarded without unrequested trimming or summaries. Model context limits still apply. Requests exceeding supported limits must be handled explicitly.
Your supported settings stay intact.
Supported reasoning settings and output limits must be honored. Unsupported options are disclosed rather than silently ignored. This is not a promise of identical responses across calls.
What comes with the offer.
- Model access
- Available versions, compatibility, supported settings and context limits.
- Pricing
- Discounted rates, billing categories, offer validity and failed-request billing.
- Usage conditions
- Region, concurrency and rate limits; any minimum spend or prepayment.
- Data handling
- Where requests go, what is logged and the applicable retention and training policies.
Who provides access?
TokenAir combines purchasing demand and buys API usage from upstream suppliers that source directly from model providers. You connect through TokenAir’s API. Multiple supply sources do not mean every model or feature is already available.