Jev
Jev is TypeSafe AI's first System One model, launched in early access on September 15, 2026. It is not a chatbot: you send unstructured or structured state plus typed questions and get structured answers your code can branch on.
Provider
TypeSafe AI
Model family
TypeSafe System One
System One decision model
Cost tier
Jev
Status
Current
Release Sep 15, 2026
Why teams choose it
Official spelling is Jev (William Stanley Jevons), not Jeva
The API model field in the quickstart is jev-latest.
TypeSafe's launch post prices input at $0.042 per million tokens ($42 per billion) and calls output too cheap to meter
Homepage workflow copy cites 193.6× faster and 444.6× cheaper versus LLM wrappers on TypeSafe's own System One workflows, using GPT-6 Astra and Claude Fable 5.1 as the reference average.
Schema matching is guaranteed: the model cannot emit an undeclared Choice option
That is not a claim that the selected option is factually correct. Access is a waitlist / early-access API key.
Tradeoffs to know
- Jev does not generate chat, code, or free-form text. JSON-mode GPT/Claude is a different job.
- Early access. TypeSafe's speed/cost headlines are vendor workflow evals, not a public gold-set bake-off.
- Choice cardinality is documented at 255 options. The Doom demo uses structured game state, not pixels.
Benchmarks
No independent or vendor benchmark scores in this catalog entry.
Technical specs
- Inputs
- text
- Outputs
- structured
- License
- Proprietary API (early access)
- API ID
jev-latest- Latency Ms Range
- 70, 500
- Choice Max Options
- 255
- Input Usd Per Million Tokens
- 0.042
- Output Usd Per Million Tokens
- 0
- Capabilities
- typed decisions
- choice classification
- scoring
- noul yes-no probability
- calibrated confidence
- parallel questions
- workflow routing
Sourcehttps://typesafe.ai/blog/introducing-system-one-models-and-jev
Jev FAQ
What is Jev?
Jev is TypeSafe AI's first System One model, launched in early access on September 15, 2026. It is not a chatbot: you send unstructured or structured state plus typed questions and get structured answers your code can branch on. The documented model alias is jev-latest on POST https://api.typesafe.ai/v1/systemone. Question primitives are Choice (up to 255 options), Score, and Noul (yes/no probability). TypeSafe lists $0.042 per million input tokens with output tokens unmetered, and 70–500 ms end-to-end on its launch workloads. Training is Reinforcement Learning for Calibrated Decisions (RLCD).
When does Jev fit best?
Classify, route, and score tickets or documents inside ordinary software
What should teams watch out for with Jev?
Jev does not generate chat, code, or free-form text. JSON-mode GPT/Claude is a different job.
Explore next
Models, tools, and comparisons that connect to this reference.