b9659

ggml-org/llama.cppb9659Jun 15, 2026by github-actions[bot]same day as b9658

What changed since b9658

low risk

b9659 is a single-commit follow-up that fixes n_tokens miscounting in multimodal (mtmd) handling.

Bug fixes

  • Corrected token counting in multimodal (mtmd) code paths that previously reported the wrong n_tokens value.

Upgrade notes

This is a one-commit patch on top of b9658 with no API, config, or build changes; binaries are otherwise identical in layout. Upgrading is only worthwhile if you use multimodal models, where the corrected token count avoids downstream token-management errors. No migration steps are required.

Analyzed Nov 1, 2026

Upgrading?

Release notes summary

Fixes a token counting error in multimodal models that could lead to incorrect token management. Standard binaries are available for all supported platforms.

Key Highlights

  • Fix token counting bug in multimodal models
  • Standard release builds for all platforms
  • Disabled KleidiAI build for macOS Apple Silicon

New Features

  • Fix token counting bug in multimodal models

Release Notes (excerpt)

<details open> mtmd: fix miscounting n_tokens (#24656) </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