b9645

ggml-org/llama.cppb9645Jun 15, 2026by github-actions[bot]same day as b9644

What changed since b9644

low risk

Adds BF16 support for the Metal repeat operation, improving Apple Silicon compatibility; no other changes noted.

New features

  • Metal backend now supports the repeat operation with BFloat16 tensors, enabling BF16 models to run that path on Apple Silicon.

Improvements

  • Improved BF16 handling on macOS/iOS Metal, which may reduce fallbacks or conversions for BF16 workloads.

Upgrade notes

This is a single-commit increment over b9644 with no config or API changes; existing flags and model files remain valid. Apple Silicon users running BF16 models on Metal are the primary beneficiaries. Non-Metal platforms see no behavioral change.

Analyzed Oct 31, 2026

Upgrading?

Release notes summary

This update adds BFloat16 (BF16) support for the Metal backend's repeat operation, enhancing performance and compatibility on Apple Silicon devices.

Key Highlights

  • Metal: Added repeat BF16 support
  • Improved BF16 handling on Apple Silicon
  • macOS and iOS binaries updated

Release Notes (excerpt)

<details open> metal : add repeat bf16 (#24638) </details> **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) **Android:** - Android arm64 (CPU) **Windows:** - Windows x64 (CPU)…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories