Rate limits
Per-organization rate limits, concurrency caps, and 429 behavior.
Limits apply per organization across all its keys.
Defaults for new accounts
- Requests: 600 requests/minute on inference endpoints.
- Concurrency: 100 in-flight inference requests.
- Spend caps: $25/day and $250/month model spend; $100/day in top-ups.
Tiers
Limits multiply with account age and lifetime spend:
| Tier | Qualifies when | RPM | In-flight | Daily spend | Monthly spend | Top-ups/day |
|---|---|---|---|---|---|---|
| 0 | new account | 600 | 100 | $25 | $250 | $100 |
| 1 | 7 days old or $10 spend (min age 1 day) | 1,200 | 200 | $100 | $1,000 | $500 |
| 2 | 30 days or $100 spend (min age 3 days) | 2,400 | 400 | $500 | $5,000 | $2,500 |
| 3 | 60 days or $1,000 spend (min age 7 days) | 6,000 | 1,000 | $5,000 | $50,000 | $10,000 |
| 4 | 90 days or $5,000 spend (min age 14 days) | 12,000 | 2,000 | $15,000 | $200,000 | $20,000 |
When limited
Over-limit requests return 429 with a JSON error. Responses include Retry-After and RateLimit-Reset (seconds until the window resets) plus RateLimit-Remaining: 0. Back off for the indicated interval and retry.
How is this guide?