TypeSafe AI Releases Jev, a System One Model for Typed Decisions in Agent Loops
TypeSafe AI released Jev, a System One model that returns typed decisions with calibrated probabilities for agent-loop tasks, and claimed 193.6x faster and 444.6x cheaper performance in its own evals.
Founder Diogo Almeida previously worked at OpenAI on the instruction-following research behind ChatGPT. The company also published 20 agentic use cases for Jev, according to its launch materials.
Every Jev call sends a state, in text or JSON, plus a dictionary of typed questions. TypeSafe's documentation defines three primitives. Choice picks one option from a list and returns a probability per option plus confidence. Score rates the state on ordered rubric levels and returns probabilities plus confidence. Noul returns the probability, from 0 to 1, that a statement is true. All questions in a request are evaluated in parallel against the same state.
TypeSafe trains Jev with Reinforcement Learning for Calibrated Decisions, or RLCD, so that higher confidence should track higher accuracy. Choice supports up to 255 options.
The performance claims come from TypeSafe's own workflow evaluations. TypeSafe says Jev is 193.6 times faster and 444.6 times cheaper. The launch post says these figures sit on the higher end of real-world gains and use GPT-6 Astra and Fable 5.1 as the reference answer.
Editor's Summary TypeSafe AI has released Jev, a System One model that returns typed decisions with calibrated probabilities rather than chat or code. It targets agent-loop judgments and offers Choice, Score and Noul primitives. TypeSafe claims large speed and cost gains from its own workflow evals using GPT-6 Astra and Fable 5.1 as reference.