llama.cpp releases b10079 with null context fix for server stability

ggml-org's llama.cpp, the widely used open-source C++ implementation for running large language models locally, has released version b10079. The release notes highlight a server-side bug fix that properly handles null llama_context (issue #25868). This fix prevents potential server crashes when the context is not initialized correctly, a critical improvement for developers running production-like inference servers on their own hardware. The commit was signed with GPG key B5690EEEBB952194, confirming its authenticity.

b10079 comes with a comprehensive set of pre-built binaries covering almost every platform. For macOS, it supports Apple Silicon (ARM64) including a KleidiAI-enabled variant, and Intel (x64). Linux builds include CPU, ARM64 CPU, s390x, and GPU-accelerated versions with Vulkan, ROCm 7.2, OpenVINO, and SYCL (FP32/FP16). Windows users get CPU (x64 & ARM64), OpenCL for Adreno, CUDA 12 & 13, Vulkan, OpenVINO, SYCL, and HIP. Mobile support includes Android ARM64, and there are UI assets for the built-in web interface. This wide compatibility ensures that llama.cpp remains the go-to engine for running LLMs locally, from development to deployment on edge devices.