llama.cpp v0.0.0-b10047 released with BoringSSL security update
New version b10047 of llama.cpp updates BoringSSL for enhanced security across all platforms.
The llama.cpp project, known for its efficient C++ implementation of Meta's Llama models, has released version b10047. This update primarily focuses on security by updating the vendored BoringSSL library to version 0.20260713.0 (commit #25624). BoringSSL is a cryptographic library forked from OpenSSL, and this update ensures compatibility with the latest security standards and fixes potential vulnerabilities. The release is signed with a verified GPG key (B5690EEEBB952194) to ensure authenticity.
Accompanying the security update, llama.cpp b10047 provides a wide array of prebuilt binaries for multiple operating systems and hardware backends. Users can download builds for macOS (both Apple Silicon and Intel), Windows (CPU, ARM64, CUDA 12 and 13, Vulkan, OpenVINO, SYCL, HIP), Linux (Ubuntu for x64/arm64 on CPU, Vulkan, ROCm 7.2, OpenVINO, SYCL FP32/FP16), Android (ARM64 CPU), iOS (XCFramework), and even openEuler with ACL Graph support. This ensures that developers and researchers can run Llama models with optimal performance and security on virtually any platform.
- llama.cpp v0.0.0-b10047 (tag b10047) released on July 16, 2025, updating BoringSSL to 0.20260713.0 for improved cryptographic security.
- Prebuilt binaries available for macOS, Windows, Linux (Ubuntu), Android, iOS, and openEuler, covering CPU, GPU (CUDA, Vulkan, ROCm, OpenVINO, SYCL), and ARM backends.
- Release verified with GPG key B5690EEEBB952194; commit 86b719b signed via GitHub's verified signature system.
Why It Matters
Keeping Llama model inference secure and accessible across platforms is critical for AI deployment in production environments.