
Jev 1.14
Jev 1.14 is the newest version of TypeSafe's decision-only AI: instead of writing text, it returns a yes/no answer, a pick from a fixed list, or a score, each with a probability. It suits software that must pick quickly among set options, such as sorting requests, routing tasks or rating risk. TypeSafe says it improves a little across the board, most in non-English languages, and adds a question type that boxes a location in an image. On our Korean and Japanese judgment tests it matched the previous Jev 1.13 and ranked near the top, and in our real-time game tests it got better at Pong but slipped a little at Tetris and Snake. TypeSafe's own docs say English is its most accurate language and that Korean, Chinese and Japanese are weaker, and the image and very long document options are still experimental and not recommended for production. The weights are not public; it is available only through TypeSafe and OpenRouter.
API cost
Real-world usage
- Classifying & tagging9.3%
- Finance & trading7.2%
- Knowledge Q&A3.7%
Expert Performance Evaluation
Classification and judgment
All classification and judgment AI rankingsSee each evaluation’s detail page for its source and last updateData sources & removal requests