b9913

ggml-org/llama.cppb9913Jul 8, 2026by github-actions[bot]same day as b9912

What changed since b9912

medium risk

b9913 adds a new 2-bit Q2_0 quantization type with CPU backend support, on top of b9912's naming cleanup.

New features

  • New Q2_0 quantization type definition enabling 2-bit precision models.
  • CPU backend implementation for Q2_0, allowing smaller model footprints on CPU-only setups.

Improvements

  • Speculative decoding module naming and spacing cleaned up in the prior release (b9912), carried forward.

Upgrade notes

Existing GGUF models and workflows are unaffected; Q2_0 only matters if you produce or load models quantized with the new type. Note that Q2_0 is CPU-only for now, so GPU offload of such models will not be accelerated. No config or CLI changes are required for existing setups.

Analyzed Oct 30, 2026

Upgrading?

Release notes summary

Introduces support for Q2_0 quantization, a 2-bit quantization scheme, specifically for the CPU backend, allowing for smaller model sizes.

Key Highlights

  • Add Q2_0 quantization type definition
  • Implement Q2_0 quantization for CPU backend
  • Allows for smaller model sizes with 2-bit precision

Release Notes (excerpt)

<details open> Add Q2_0 quantization: type definition and CPU backend (#24448) </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