b9030
The open-source LLM inference engine now supports 20+ platform/backend combos with a security update.
ggml-org has released llama.cpp b9030, a stable update to the widely-used open-source C++ inference engine for large language models. The headline change is the vendor dependency update of cpp-httplib to version 0.43.3, which brings improved security, bug fixes, and better performance for HTTP-based operations. While no new model architectures or major features were added, this release prioritizes reliability and compatibility for the thousands of developers who run LLMs locally via llama.cpp.
The most notable aspect of b9030 is the sheer breadth of prebuilt binaries provided. The release includes builds for macOS (Apple Silicon with and without KleidiAI, Intel, iOS XCFramework), Linux (x64 CPU, arm64 CPU, s390x, x64 with Vulkan, arm64 with Vulkan, x64 with ROCm 7.2, OpenVINO, SYCL FP32/FP16), Windows (x64 CPU, arm64 CPU, x64 with CUDA 12/13 with required DLLs, Vulkan, SYCL, HIP), Android (arm64 CPU), and openEuler (x86 and aarch64 with Ascend hardware support). This extensive platform support allows developers to deploy LLM inference on almost any modern device or server, from desktops to edge devices, without needing to compile from source. With 109,000 GitHub stars and 17,800 forks, llama.cpp continues to be the dominant open-source engine for local AI inference.
- Updated cpp-httplib to 0.43.3 for enhanced HTTP handling and security fixes
- Prebuilt binaries for over 20 platform/backend combinations including CPU, CUDA, Vulkan, ROCm, OpenVINO, SYCL, and HIP
- 109k stars and 17.8k forks on GitHub, reflecting massive community adoption
Why It Matters
For developers deploying LLMs locally, b9030 ensures a stable, secure, and cross-platform inference engine with zero compilation hassle.