ollama v0.32.3: Improved GPU Support and Model Downloads
▼ WHAT HAPPENED
ollama v0.32.3 introduces several critical improvements for local AI operators running on consumer hardware. The release fixes model download stalls, enhances integrations with Claude Code Channels and Anthropic thinking streams, and ensures Hermes Desktop respects the `--force-build` flag. Notably, it expands GPU support to include CUDA on Windows ARM64 and B200 through CUDA 12, while also reducing memory usage on Linux iGPUs. Additionally, chat, thinking, and tool calling support for Laguna 2.1 models is added with a Metal inference fix. Operators should update their configurations to leverage these enhancements and ensure compatibility with the latest model features.
▼ OPERATOR ANGLE
Upgrade to ollama v0.32.3 to benefit from improved GPU support and fixed integrations; test your setup thoroughly before deploying in production.