Limits and billing
Project limits may differ from the defaults below. These are not SLAs, plan names, or guaranteed pricing.
Common defaults
| Limit | Typical default |
|---|---|
| Requests per minute | Commonly 60 |
| Concurrent requests | Commonly 10 |
| Monthly spend | When enabled → spending_limit_exceeded |
| Request body size | 2 MB (413) |
| Default max output | 4096 tokens if neither max_tokens nor max_completion_tokens is set |
| Gateway timeouts | Finite; longer for reasoning workloads |
On 429, honor Retry-After and rate-limit headers — see Errors.
Cost notes
Note
Failed upstream attempts may still incur cost. A compatible fallback may create an additional attempt. See Models and aliases.