TypeSafe AI Introduces System One Models and Jev
TypeSafe AI, founded by ex-OpenAI researcher Diogo Almeida, has announced System One Models — a new class of AI models built for fast, structured decision-making rather than chat. The company’s first public model, Jev, is now available in early access and achieves comparable intelligence to frontier LLMs on structured tasks while being up to 200x faster and 444x cheaper.
Unlike traditional LLMs that generate text token-by-token, Jev outputs type-safe structured values with calibrated confidence scores in a single parallel query. The model uses a novel training method called Reinforcement Learning for Calibrated Decisions (RLCD) and is designed to slot into production software as a “fuzzy decision rule” — classify, route, score, or extract where hand-written logic is too brittle.
The release includes workflow evaluations and interactive demos, including a real-time Doom bot and a wikiracing game that showcase the model’s sub-500ms response times.