GitSeonjeap
English

← News

Local AI

llama.cpp b12441 Pre-release Fixes WebGPU Mask Bug

Posted As of

llama.cpp pre-release b12441 (2026-11-16) fixed a bug where the WebGPU flash-attention shader read the mask without its head dimension. b12436, released the same day, fixed a Vulkan 32-bit compilation issue.

What happened

Source: ggml-org/llama.cpp GitHub releases page

  • Pre-release b12441 was tagged on 2026-11-16 (official)
  • The WebGPU flash-attention shader read the mask without a head dimension, which the release notes say produced incorrect results when the mask had two or more heads
  • The fix adds the head index to the mask offset
  • Includes a change that makes Vulkan top-k tie selection deterministic. The release notes explain that the radix select using an atomic counter could leave different columns depending on thread scheduling
  • b12436 (2026-11-16) fixed a Vulkan 32-bit compilation issue

Our take

For users running local models on PC through the Vulkan or WebGPU paths, this fix concerns attention result accuracy and result reproducibility. However, the list of affected models, performance figures, and how much output differed in earlier builds have not been confirmed. Any direct impact on Mac or glasses devices is not confirmed in the notes, and the individual tag pages were not opened.

Sources and related posts

More in Local AI

  1. NVIDIA Releases Nemotron for Science for Scientific Agents (nvidianews.nvidia.com)

    NVIDIA announced a package based on Nemotron 3.5 Super VL. It includes a chemistry dataset and a post-training recipe, and the IR copy states that the model is optimized to run locally on a 128GB DGX Spark.

    Local AI| |1 sources |Our take included
  2. NVIDIA RTX Spark Compact Desktops Begin Shipping (blogs.nvidia.com)

    Compact RTX Spark desktops from ASUS, Dell, and MSI have launched as of the announcement date and are positioned for local AI agents and model fine-tuning. Pricing was not confirmed on NVIDIA's page.

    Local AI| |1 sources |Our take included
  3. llama.cpp b12497 adds tiny-voca support (github.com)

    llama.cpp pre-release b12497 adds tiny-voca support to the multimodal path and improves G2P conversion. The release page has no performance figures or description of the target model.

    Local AI| |1 sources |Our take included

All news