b10214

ggml-org/llama.cppb10214Jul 31, 2026by github-actions[bot]same day as b10213

What changed since b10213

low risk

b10214 adds n_embd_head support to the multimodal (mtmd) path, a single-commit follow-up to b10213's rotated KV cache quantization.

New features

  • Multimodal (mtmd) models now expose and use n_embd_head, improving handling of vision/audio embedding dimensions.

Improvements

  • Multimodal model plumbing refactored to carry head-embedding size alongside existing embedding parameters.

Upgrade notes

This is a one-commit increment over b10213, so no config or CLI changes are expected; the rotated KV cache quantization from b10213 remains intact. Rebuild or pull the new prebuilt binaries if you rely on multimodal (mtmd) models, since the change touches that code path. No migration steps are required.

Analyzed Nov 1, 2026

Upgrading?

Release notes summary

Adds n_embd_head support to multimodal models and includes builds for various platforms including macOS, Linux, Windows, and Android.

Key Highlights

  • Multimodal model support enhancement (n_embd_head)
  • Vulkan backend support for multiple platforms
  • CUDA 12 and 13 support for Windows x64
  • OpenVINO and SYCL support for Linux x64
  • UI build included

Release Notes (excerpt)

<details open> mtmd: add n_embd_head (#26342) Co-authored-by: Daniel Han <unslothai@gmail.com> </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