Back to events
Open-source releaseHigh global significanceConfirmed confidence

Kev 1.0 pins a four-size family of Jev-style open decision models

Kev 1.0 pins checkpoints, model cards and evaluations for 0.8B, 4B, 9B and 27B decision models compatible with TypeSafe's System One API.

Event details

Released on October 1, Kev 1.0 versions the existing 0.8B, 4B, 9B and 27B checkpoints, cards, evaluation suites and serving code. No models were newly trained for this versioned release. The self-hostable project continues its Jev-inspired direction with a TypeSafe-compatible interface. Its release documents evaluation scope and limitations; the results do not establish universal superiority over Jev.

Why it matters

Provides a reproducible, multi-size open-source baseline and deployment alternatives for Jev-style decisions.

70/100Global significance score. Regional effects are recorded only when the evidence supports a meaningful difference.

Access notes

Code and weights or adapters are public; the 27B full weights are hosted on Hugging Face.