v0.21.1-rc1
📦 ollamaView on GitHub →
✨ 1 features🐛 3 fixes🔧 1 symbols
Summary
This release introduces kimi CLI integration and includes several performance and correctness fixes across MLX models and server formatting logic.
✨ New Features
- Added kimi CLI integration with the installer flow.
🐛 Bug Fixes
- Repeat penalties are now applied in the sampler for MLX models.
- Sigmoid router head is fused in glm4_moe_lite for MLX.
- Format is now applied when think=false with a thinking-capable parser in the server.