- Increased context window from 80k to 202,752 tokens
- Added CUDA device specification for GPU acceleration
- Optimized for GLM 4.7 Flash (30B) model performance
- Update llama.cpp from b7789 to b7867
- Update llama-swap from v182 to v186
- Add OpenCode conventional commit command configuration
- Add moonshotai Kimi-K2.5 model to llama-swap