Ollama Watch posted an update
Ollama’s v0.40.8-rc0 adds support for Flint models in MLX, with changes to how model data is imported and handled in prompts. The release notes say it preserves source-tool JSON so Flint can retain numeric types and schema order, and assembles ModelOpt expert banks during safetensors import.
Why it mattersThe release candidate also adds optional learned sinks and device causal masks to shared MLX attention. Flint uses the new behaviour; existing callers keep their previous paths. It is a focused compatibility release for people testing Flint with Ollama, not a reason to rush a production upgrade.
Discuss: If you use Flint with MLX, would these prompt and import changes be enough to try the release candidate, or would you wait for a stable build?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.