Jev is quickly beocming the primary choice for classification requests on OpenRouter. It took 27% of weekly request volume in this category, nearly 2x the share of DeepSeek V4 Flash which previously held the top spot

Jev is a new kind of model: fast, cheap, and accurate enough to make judgment calls. You ask it a question about text or structured data and it returns a probability. Is this email urgent? Does this paragraph contain a particular writing tic? @hammer_mt used it to build Jevgram, a tool that scores text against questions such as “Does this sound like AI?” His suggestion: Give a coding agent more specific checks—like looking for “not X but Y” constructions—so it can revise flagged passages as it writes rather than after. What makes it useful is that the feedback is cheap enough to run constantly, not just at the end. Start with a judgment you could make in a few seconds, define what a useful answer looks like, and test whether Jev agrees with you on a handful of examples before running it across the whole collection. Read the full piece:

Jev Router keeps a model that works for the rest of the session. It can raise or lower effort without switching models. It switches only when the expected gain is larger than the cost, including the cached chat it would lose.

Jev reads the conversation text only to pick the model and effort. It runs under zero data retention (ZDR) terms, so nothing is stored or trained on. Attachments are never sent to Jev, and requests with "zdr: true" work with Jev Router.

Jev Router runs on Jev, TypeSafe's first Decision / System One model. Before each turn, Jev reads your prompt and scores it on difficulty and precision. Also checks whether a bigger model or more effort would help, whether a cheaper model is enough, and whether the task changed.

Jev for RAG in almost all cases you trust the semantic matching capability of Jev more than dot product similarity very useful as the direct similarity metric in small data cases and a great reranker with big data