What Is Jev? A Guide to TypeSafe AI’s System One Model

TypeSafe AI has released Jev, a model the company describes as a System One model. According to a LangChain guide, Jev does not generate text; it evaluates a state and returns typed answers and probabilities, trained with reinforcement learning for calibrated decisions (RLCD). TypeSafe AI reports up to 200x faster inference and 400x lower cost than comparable LLMs on classification tasks.

The model accepts a state and questions about that state in a single request. Three question types are supported: Choice, which returns a probability for each option plus an overall confidence score; Score, which rates input against ordered levels and returns a continuous score, a distribution and a confidence value; and Noul, a yes-or-no question returning the probability a statement is true. Multiple questions about the same state are evaluated in parallel, and adding questions barely changes response time while costing only tokens for the extra questions.

LangChain exposes Jev through TypeSafeClassifier in its langchain-typesafe integration. Developers install the package, set a TYPESAFE_API_KEY, and pass state and questions to .invoke() to receive classification results. State can be text, structured data, or LangChain messages.

The guide presents Jev as a complement to a driving LLM rather than a drop-in replacement, with example middleware for model routing and for checking tool calls before execution via AutoModeMiddleware. LangChain also cites Kyle Jeong of Browserbase powering browser use agents, Jarrod Watts building a live trading agent, and Ryan Vogel doing email triage at scale.