llama.cpp releases b10446 with BoringSSL update and expanded platform support

llama.cpp, the ubiquitous C/C++ inference library for large language models, has dropped a new release tagged b10446. The headline change is a vendor update that bumps BoringSSL to version 0.20260813.0 (PR #27099). BoringSSL is Google's fork of OpenSSL, used to provide secure TLS connections; updating it helps patch known vulnerabilities and ensures compatibility with modern cryptography. While this isn't a headline-grabbing feature like a new quantization format, it's a crucial maintenance release for anyone running llama.cpp in production or exposed to the internet. The tag was released by the GitHub Actions bot and signed with a verified GPG key, continuing the project's security-conscious release pipeline.

The new build expands accessibility with precompiled binaries for a staggering range of targets. Developers can grab versions for macOS Apple Silicon (including a KleidiAI-enabled build) and Intel, iOS XCFramework, Linux on x64/arm64/s390x, Windows on x64/arm64, and Android arm64. GPU acceleration is covered across CUDA 12 and 13, Vulkan, ROCm 7.14, OpenVINO, SYCL, and OpenCL for Adreno. There are also specialized openEuler builds with Ascend 310p and 910b support. With over 124k stars and 21.7k forks, llama.cpp remains the go-to for developers deploying LLMs on edge devices, laptops, and data centers. This b10446 release reinforces the project's commitment to security and broad hardware support, making local AI inference more robust than ever.