Jev rate limits
The models page publishes:
- 250,000 tokens per second
- 1,200 requests per minute
Exceeding either returns 429 Too Many Requests. TypeSafe also documents 529 Overloaded when their side is temporarily overloaded.
They are explicit that limits are adjusting dynamically under demand and “can change without notice.” Higher limits: custom and enterprise, sales@typesafe.ai.
Client behavior
Official SDKs retry with backoff by default and honor retry-after when present. If you call HTTP yourself, implement exponential backoff. Do not tight-loop 429s.
Design around the request budget
Batching questions reduces request-rate pressure and repeats the state once. Fan-out of HTTP calls (one question each) is how you hit 1,200 RPM without meaning to.
We do not publish unofficial “hidden” limits.
Sources
Public TypeSafe or adjacent documentation only. No private claims.