Ollama v0.40.0 switches models to MLX on Apple Silicon by default

The release Ollama v0.40.0 is out. The main change: on Apple Silicon devices, models whose architectures are supported by the MLX runtime now automatically run through MLX. The developers' example — ollama pull qwen3.8 and ollama run qwen3.8. During the pre-release, the authors promise to test and enable additional models.

What it means

For those running models on a Mac with M-series chips, this is a change in default behavior: no separate setting is needed to switch to MLX. If a model's architecture is supported by MLX, Ollama will pick that runtime on load. The developer does not specify which architectures are already enabled and which will come later — the release notes only state that the list will expand during pre-release testing. The update is worth tracking for those who work with models on Apple Silicon and want to understand how performance and memory consumption change on their specific tasks. Since the release is marked as pre-release, behavior may change from build to build. You can follow news about tools for running neural networks locally in the CPU3D news feed.