b8531
The latest commit prevents data loss during cache updates and adds new Windows CUDA 13.1 DLLs.
The open-source project llama.cpp, maintained by ggml-org, has released a significant update with commit b8531. This release addresses a critical issue (#21000) where the system was incorrectly deleting old cache files during software updates, which could lead to data loss and require re-downloading of model weights. The fix ensures smoother updates for the millions of developers using llama.cpp to run Llama, Mistral, and other GGUF-format models locally on consumer hardware.
The update also expands hardware support with new Windows CUDA 13.1 DLLs, giving NVIDIA GPU users more driver flexibility. This continues llama.cpp's mission of providing the most accessible local AI inference, maintaining compatibility across macOS (Apple Silicon and Intel), Linux (CPU, Vulkan, ROCm 7.2, OpenVINO), Windows (CPU, CUDA, Vulkan, SYCL, HIP), and openEuler systems. The project now boasts 99.4k GitHub stars, reflecting its crucial role in the democratization of AI model deployment.
- Fixes critical bug #21000 preventing deletion of old cache files during updates
- Adds Windows CUDA 13.1 DLL support for NVIDIA GPU users with newer drivers
- Maintains broad platform support across macOS, Linux, Windows, and openEuler systems
Why It Matters
Prevents data loss for developers running local AI models and expands hardware compatibility for broader adoption.