b10262

ggml-org/llama.cppb10262Aug 4, 2026by github-actions[bot]same day as b10261

What changed since b10261

medium risk

Adds Vulkan backend support for the GATED_LINEAR_ATTN operation, with no changes outside that backend.

New features

  • GATED_LINEAR_ATTN is now implemented in the Vulkan backend, enabling gated linear attention models to run on Vulkan GPUs.

Improvements

  • Vulkan operations documentation (ops.md) was refreshed to reflect the newly supported op.
  • An unused GLA specialization constant was removed from the Vulkan shader code, simplifying the pipeline setup.

Upgrade notes

This is a single-commit increment over b10261; the only functional delta is additive Vulkan op coverage, so no config, model, or CLI changes are required. Users on Vulkan who previously fell back to CPU for gated linear attention models should see those graphs execute on-device after upgrading. Non-Vulkan backends are unaffected.

Analyzed Oct 31, 2026

Upgrading?

Release notes summary

Implements the GATED_LINEAR_ATTN operation for the Vulkan backend.

Key Highlights

  • Vulkan backend support for GATED_LINEAR_ATTN
  • Updated documentation for Vulkan ops
  • Removed unused GLA spec constant

Release Notes (excerpt)

<details open> vulkan backend ops: implemented GATED_LINEAR_ATTN (#25601) * vulkan : add GATED_LINEAR_ATTN op * docs : update Vulkan ops * vulkan : remove unused GLA spec constant * Updated ops.md * ops.md update </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)…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories