b10436

ggml-org/llama.cppb10436Aug 15, 2026by github-actions[bot]same day as b10435

What changed since b10435

low risk

b10436 follows b10435's jinja performance fix with small mtmd/common bug fixes and a reverted gguf change.

Improvements

  • Carries forward the jinja gather_string_parts optimization from b10435 that removed quadratic string-processing cost.
  • Bundles assorted stability fixes across the mtmd (multimodal) and common modules.

Bug fixes

  • Addresses several defects in the mtmd module affecting multimodal handling.
  • Applies miscellaneous correctness fixes in the common module.
  • Reverts a prior gguf fix, restoring the earlier behavior to avoid regressions.

Upgrade notes

This is a small incremental step over b10435 with no API or configuration changes expected. Note that a gguf-related fix was reverted, so if you depended on that behavior in an intermediate build, re-validate gguf loading. Multimodal users should re-test mtmd paths after upgrading.

Analyzed Oct 24, 2026

Upgrading?

Release notes summary

Various bug fixes for multi-token and common modules, addressing stability issues without introducing new features.

Key Highlights

  • Bug fixes for mtmd module
  • Common module improvements
  • Reverted gguf fixes to maintain stability

New Features

  • Bug fixes for multi-token and common modules

Release Notes (excerpt)

<details open> mtmd, common: various fixes (#27071) * apply fixes * cont * revert gguf fix </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 7.14)DISABLED - Ubuntu x64 (OpenVINO) - Ubuntu x64 (SYCL FP32)…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories