v0.32.3
Added
- Added CUDA support on Windows ARM64
- Added B200 GPU support through CUDA 12
- Added chat, thinking, and tool calling support for Laguna 2.1 models
Changed
- Made Hermes Desktop respect --force-build flag
- Reduced memory usage on Linux CUDA/ROCm iGPUs
- Updated the MLX engine
- Updated the llama.cpp engine
Fixed
- Fixed model downloads that stall before sending data
- Restored Claude Code Channels in integrations
- Fixed Anthropic thinking streams in integrations
- Fixed Metal inference for Laguna 2.1 models
- Fixed GLM tool calls being silently dropped at the end of generation
What's Changed
- Fixed model downloads that stall before sending data.
- Improved integrations: restored Claude Code Channels, fixed Anthropic thinking streams, and made Hermes Desktop respect
--force-build. - Expanded GPU support with CUDA on Windows ARM64, B200 support through CUDA 12, and lower memory use on Linux CUDA/ROCm iGPUs.
- Added chat, thinking, and tool calling support for Laguna 2.1 models, including a Metal inference fix.
- Fixed GLM tool calls being silently dropped at the end of generation.
- Updated the MLX and llama.cpp engines.
Full Changelog: https://github.com/ollama/ollama/compare/v0.32.1...v0.32.3