Jev pricing
Jev is priced for high-volume, machine-scale use: you pay only for input, and output is free.
What that actually means
At $0.042 per million input tokens, Jev is one to two orders of magnitude cheaper than calling a frontier LLM for the same classify/route/score task — and it returns a typed answer instead of text you'd have to parse. TypeSafe estimates an average decision costs about $0.0004 on their benchmarks.
- 1,000 decisions ≈ $0.40
- 1 million decisions ≈ $400
- Output tokens are free, so batching several questions into one call is cheap.
Per-decision figures are estimates from TypeSafe's published benchmarks; your real cost depends on how large your state is, since you pay per input token. Always confirm current pricing on the official site.
Rate limits
- 250,000 tokens / second
- 1,200 requests / minute
How to lower your bill
- Trim the state. You pay per input token — send only the context the decision needs.
- Batch questions. Several questions share one state payload in a single round trip.
- Route by confidence. Let Jev auto-handle the easy, high-confidence cases and reserve expensive LLM calls or human review for the uncertain few.
Getting access
Official keys, quotas and billing are managed by TypeSafe at console.typesafe.ai. You can try the model right now — no key, no waitlist — in our free playground.