b9654

ggml-org/llama.cppb9654Jun 15, 2026by github-actions[bot]same day as b9653

What changed since b9653

medium risk

b9654 adds a post-decode callback hook for multimodal (mtmd) models; otherwise identical to b9653.

New features

  • Multimodal (mtmd) pipeline now exposes a post-decode callback so callers can run custom processing after the decode phase.

Improvements

  • Release binaries rebuilt for all standard platforms (macOS, iOS, Linux CPU/Vulkan/ROCm/OpenVINO/SYCL, Windows, Android, openEuler) with no packaging changes.

Upgrade notes

This is a single-PR delta over b9653, so no config or model migration is required. Only consumers of the mtmd/multimodal C/C++ API need to rebuild to take advantage of the new post-decode callback; CLI users can upgrade in place. The KleidiAI macOS Apple Silicon build remains disabled, as in the previous release.

Analyzed Nov 1, 2026

Upgrading?

Release notes summary

Adds a post-decode callback for multimodal models, allowing custom processing steps after the decoding phase. Standard binaries are provided for all supported operating systems.

Key Highlights

  • Add post-decode callback for multimodal models
  • Standard release builds for all platforms
  • Disabled KleidiAI build for macOS Apple Silicon

Release Notes (excerpt)

<details open> mtmd : add post-decode callback (#24645) Assisted-by: pi:llama.cpp/Qwen3.6-27B </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:**…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories