b10066

ggml-org/llama.cppb10066Jul 17, 2026by github-actions[bot]same day as b10064

What changed since b10064

low risk

Adds MoE support to OpenCL using binary kernel library

New features

  • Adds MoE (Mixture of Experts) support for OpenCL devices
  • Loads specific kernels from binary kernel library

Improvements

  • Improves performance for MoE models on OpenCL
  • Updates kernel loading mechanism to use binary library

Bug fixes

  • Fixes issue #25797 regarding OpenCL kernel loading

Upgrade notes

No configuration changes are required. Ensure binary kernel files are present for MoE support.

Analyzed Oct 16, 2026

Upgrading?

Release notes summary

OpenCL update to load and use a specific kernel from binary kernel library.

Key Highlights

  • Load and use kernel_gemm_moe_q6_k_f32_ns from bin kernel lib
  • Improves MoE (Mixture of Experts) support on OpenCL
  • Fixes issue #25797

Release Notes (excerpt)

<details open> opencl: load and use `kernel_gemm_moe_q6_k_f32_ns` from bin kernel lib (#25797) </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)…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories