Jev is an AI model from TypeSafe AI, an AI lab. TypeSafe calls it a System One Model: a class of model built to make decisions inside software, not to chat. A large language model (LLM) writes text one token at a time; Jev instead receives a structured question with unstructured data, such as text, and returns typed decisions. The possible answers are defined in advance, and each answer comes with a calibrated probability and a confidence score.
Because the type of every answer is fixed, the calling code uses the result directly, without parsing, and sets thresholds for when to act alone and when to ask a person for review. Jev computes all outputs in parallel in one query; TypeSafe states response times of 70 to 500 milliseconds. Typical uses are to classify, route, score and extract data, and to branch inside a workflow.
Jev is a hosted API in early access. Use is billed per input token; output tokens are free.