Jev FAQ: Common Questions About TypeSafe's Jev
Last checked · Independent guide, not affiliated with TypeSafe AI
Jev is TypeSafe AI's decision model: it answers typed questions about text with probabilities instead of writing text. It costs $0.042 per million input tokens with free output, launched on September 15, 2026 in early access, and is closed source. Each question below links to a page with the full answer.
Each question gets a short answer here and a full page of its own. Answers reflect what was true on September 19, 2026.
The basics
Section titled “The basics”What is Jev? A model from TypeSafe AI that returns typed decisions (a yes/no probability, a choice from a list, or a score on your scale) about text you send it. It does not write text. What is Jev?
Is Jev an LLM? Not in the usual sense. It understands language like an LLM and is reported to be transformer-based, but it never generates text and is trained for calibrated decisions. Is Jev an LLM?
Why is it called Jev? After the economist William Stanley Jevons, whose paradox says cheaper resources get used more. Why is it called Jev?
Who makes Jev? TypeSafe AI, a San Francisco startup led by former OpenAI researcher Diogo Almeida. TypeSafe AI
Access and cost
Section titled “Access and cost”How do I get access? Through TypeSafe’s console (early access, with a waitlist) or, without approval, through OpenRouter, Vercel AI Gateway or Cloudflare Workers AI. How to get a Jev API key
Why is there a waitlist? Capacity: TypeSafe says it is adjusting limits while it adds GPU capacity for launch demand. Jev waitlist
Is Jev free? Not on TypeSafe or OpenRouter. Vercel listed it as free under a promotion ending September 25, 2026, with a card required. Is Jev free?
How much does it cost? $0.042 per million input tokens, output free. A typical request costs about $0.00002. Jev pricing
What are the rate limits? 1,200 requests per minute and 250,000 tokens per second on TypeSafe’s API, adjusted dynamically; Vercel’s free tier is much stricter. Jev rate limits
Capabilities
Section titled “Capabilities”Can Jev hallucinate? It cannot return an answer outside your options, but it can pick the wrong option. Can Jev hallucinate?
Can Jev write text or code? No. It decides; an LLM writes. Can Jev generate text?
What is it good for? Classification, routing, moderation, guardrails, ranking and agent decisions. What can Jev do?
What is it bad at? Math, counting, dates, multi-step reasoning and long noisy inputs. Jev limitations
How does it compare with GPT, Claude or Gemini? Far cheaper and faster for narrow decisions; no use for writing or deep reasoning. Jev vs LLMs
Openness and data
Section titled “Openness and data”Is Jev open source? No. There are no open weights or paper; the SDKs and an agent skill are open source. Is Jev open source?
Can I fine-tune Jev? No. You customize it through the state, instructions and criteria. Can you fine-tune Jev?
Does Jev train on my data? TypeSafe says it does not train on customer requests or responses. Does Jev train on your data?
Troubleshooting
Section titled “Troubleshooting”Is Jev down? Check status.typesafe.ai; on September 19, 2026 it showed 99.854% availability over 90 days. Is Jev down?
I got an error. What does it mean? Every error code we have seen, with fixes: Jev API errors