b9692

ggml-org/llama.cppb9692Jun 17, 2026by github-actions[bot]same day as b9691

What changed since b9691

low risk

b9692 fixes LLaVA-UHD multimodal handling by removing the batch dimension, with no other changes noted.

Bug fixes

  • LLaVA-UHD multimodal model no longer uses the batch dimension, correcting a dimensional inconsistency in mtmd.

Upgrade notes

This is a single-commit follow-up to b9691, so upgrading is straightforward with no config or API changes. Rebuild/redeploy if you use LLaVA-UHD vision models, since the fix affects multimodal input handling. The POWER11 build gating from b9691 remains in place.

Analyzed Nov 1, 2026

Upgrading?

Release notes summary

Fixes the LLaVA-UHD multimodal model to no longer use the batch dimension, correcting a dimensional inconsistency.

Key Highlights

  • LLaVA-UHD batch dimension correction
  • Multimodal model fix

New Features

  • LLaVA-UHD batch dimension fix

Release Notes (excerpt)

<details open> mtmd: llava_uhd should no longer use batch dim (#24732) </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:**…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories