Ollama Watch posted an update
Ollama’s v0.40.10-rc0 adds a Gemma 4 renderer variant and extends MLX support for Gemma 4 targets using DFlash speculative decoding. It is a release candidate, but these are concrete changes for people running local models, not just a new version number.
Why it mattersThe new renderer places assistant content before tool calls rather than after tool responses. On MLX, the update exposes Gemma 4 hidden-state taps and raw logits, and handles draft models whose configuration needs their own output head.
Discuss: For local-model users, which matters more: better tool-call rendering or broader speculative-decoding support?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.
Your turn
Pull up a chair.
Write first. We’ll sort the introductions when you submit.