Kev: tiny Jev-like decision models (0.8B/4B/9B) on Qwen3.5 you can train and run locally - the 9B fits a 32GB Mac

Been playing with judge models in my eval pipeline, so this landed at the right time. Kev is a small family of Jev-architecture decision models (0.8B, 4B, 9B) built on Qwen3.5, Apache-2.0, with training code and frozen eval suites included. The 9B fits a 32GB Mac in bf16, and it's API-compatible with TypeSafe's System One, so their Python SDK points at your local server with no changes. The web playground is worth a look on its own, it shows how option order shifts the answers, which is the bias everyone building LLM-as-judge runs into. Repo: https://github.com/jaredpalmer/kev