On 25 September, Ollama released version 0.40.0-rc0. The letters “rc” mean it is a pre-release being tested before a stable launch.
On Macs with Apple silicon, such as M1, M2 and later chips, Ollama will now use MLX automatically for supported models. MLX is Apple’s open-source machine learning framework for its own chips. Previously, choosing the technology behind a local run could require more knowledge or special versions.
This does not mean every model suddenly becomes faster. Support still covers only some model types, and Ollama says more will be tested during the pre-release period. If you depend on a stable installation, there is no need to rush this update.
The interesting part is the direction: local AI becomes more useful when the software can choose a suitable engine by itself. Less time spent on settings leaves more time to compare models, write and create on your own computer.