Jev

Jev is the first publicly available model from TypeSafe AI, a San Francisco company founded in 2024 by Diogo Almeida, who co-invented the RLHF and InstructGPT methods behind ChatGPT while at OpenAI. Instead of generating text, Jev takes a block of unstructured state plus a set of typed questions and returns structured answers, each carrying a calibrated probability and a confidence value. TypeSafe calls this class of model a System One Model, trained with a method it names Reinforcement Learning for Calibrated Decisions. The intended user is a developer placing a judgement call inside ordinary code: classifying, routing, scoring, extracting, or branching where hand-written rules are too brittle. Because the output schema is fixed in advance, the model cannot return a type error, so code can act automatically above a chosen confidence threshold and escalate below it. Jev is served from a single endpoint with Python and JavaScript SDKs, priced at $0.042 per million input tokens with output tokens free. The trade-off is deliberate. Jev writes no text at all, so it cannot chat, draft, or produce code, and it accepts text input only. It is in early access behind a waitlist, and the headline speed and cost comparisons published alongside it are the company's own benchmarks.