TypeSafe’s Jev is a System One decision model: you send a state plus typed questions, and you get structured answers back — a Choice, a Score, or a Noul probability — not chat prose.
If you only need “which tool runs next” or “how bad is this ticket,” parsing JSON out of a chat model is slow and fragile. Jev is built for that gap.
What you can do today without the vendor waitlist
On jev-agent.com you can:
- Get a free API key (email, no card) for eight tuned rulesets and call decisions without writing Jev question shapes yourself
- Run live demos in the browser: chess and wikirace
- Read use-case writeups with live-API measured cost and latency: triage, tool selection, moderation, routing, and more
Numbers from live calls (not a slide deck)
Rate card: $0.042 per 1M input tokens, output free (no output tokens).
Examples reported from live API calls:
- One Choice over 3 options: $0.0000144 (342 input tokens)
- Choice + Score + Noul in one call: $0.0000226 (539 tokens)
Latency on the hot path is typically 70–500ms end-to-end depending on route. Putting several questions in one request matters: they share one state and run in parallel, so you don’t pay for the state three times.
When it fits
Reach for Jev when the answer is a decision, not a sentence: routing, classification, tool pick, rubric grading, boolean guardrails. Skip it when you need prose, code, or a conversation — pair it with an LLM for that half.
Not affiliated
jev-agent.com is an independent guide with measured figures and demos. Not affiliated with TypeSafe AI. Where this site and the official docs disagree, trust docs.typesafe.ai.
Start here: https://jev-agent.com · free key: /api-access · cases: /use-cases · pricing: /pricing
Top comments (0)