The engineering behind inference.Contact ↗
Batch & Release/Inference

Ollama v0.40.4-rc1 adds Kev decision scoring to MLX backend

Ollama's release candidate v0.40.4-rc1 introduces Kev decision scoring to the MLX backend, loading Kev models with FP32 decision heads and release metadata, plus System One scoring with shared-state caching.

Ollama v0.40.4-rc1 adds Kev decision scoring to MLX backend
Illustration generated with AI for this article.

Ollama tagged v0.40.4-rc1 on October 13, a release candidate that adds Kev decision scoring to its MLX backend. The change, merged as pull request 18976, loads Kev models with their FP32 decision head and their release metadata, and supports System One scoring with shared-state caching and limits specific to Kev models.

What changes for MLX users

The MLX backend, which runs Ollama on Apple Silicon, can now load Kev models together with their decision head, kept in FP32 precision, and their release metadata. Scoring in System One mode reuses shared state through a cache, and Kev models run under limits of their own.

Release status

The tag is a release candidate, not a stable release, so its behavior can still change before the final version. Teams that want Kev decision scoring on MLX should test the candidate before relying on it in production.

Sources & references

Ollama: v0.40.4-rc1: Add Kev decision scoring to MLX (#18976) ↗Source consulted: 2026-10-14

Privacy choices

We remember your answer in this browser for 180 days. Come back here whenever you want to change it.

EssentialSaves your answer in this browser. No cookies, no third parties.Always active