v0.16.2-rc0: mlxrunner fixes (#14247)
A new release quietly fixes major issues for running AI models on Macs.
Deep Dive
Ollama has released version v0.16.2-rc0, a release candidate focused on fixing bugs in its MLX runner. The update specifically addresses loading issues for the GLM-4-MoE-Lite model and diffusion models, removes extraneous log lines, and fixes the `--imagegen` flag. This patch is crucial for developers using Apple Silicon (M-series chips) to run local AI models, ensuring stability and correct functionality for popular architectures on macOS hardware.
Why It Matters
This ensures Mac users can reliably run cutting-edge open-source models like GLM-4 locally without crashes or errors.