b10278

ggml-org/llama.cppb10278Aug 5, 2026by github-actions[bot]same day as b10276

What changed since b10276

low risk

Build-script-only delta: the GGML_METAL_USE_BF16 CMake flag is removed, with no runtime or API changes.

Breaking changes

  • Custom builds or CI pipelines that still pass -DGGML_METAL_USE_BF16=ON will get a CMake unused-variable warning and the flag will silently have no effect.

Improvements

  • Metal build configuration is simplified by dropping the GGML_METAL_USE_BF16 option from every build script.
  • Reproducible, security-hardened dependency installation (npm ci) from the prior release remains in place.

Upgrade notes

Strip GGML_METAL_USE_BF16 from CMake invocations, CI matrices, and packaging scripts before bumping the pin. Prebuilt binaries behave the same as b10276, so no model or runtime migration is required. Re-check Metal throughput after upgrading, since bf16 handling is now decided internally rather than by the flag.

Analyzed Oct 31, 2026

Upgrading?

Release notes summary

Removes the GGML_METAL_USE_BF16 build flag from all build scripts.

Key Highlights

  • Removes GGML_METAL_USE_BF16 from all build scripts

Release Notes (excerpt)

<details open> build : remove GGML_METAL_USE_BF16 from all build scripts (#26604) </details> **Website:** - <https://llama.app> **macOS/iOS:** - macOS Apple Silicon (arm64) - macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED - macOS Intel (x64) - iOS XCFramework **Linux:** - Ubuntu x64 (CPU) - Ubuntu arm64 (CPU) - Ubuntu s390x (CPU) - Ubuntu x64 (Vulkan) - Ubuntu arm64 (Vulkan) - Ubuntu x64 (ROCm 7.2) - Ubuntu x64 (OpenVINO) - Ubuntu x64 (SYCL FP32) - Ubuntu x64 (SYCL FP16)…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories