b10213

ggml-org/llama.cppb10213Jul 31, 2026by github-actions[bot]same day as b10212

What changed since b10212

medium risk

Adds quantization support for rotated KV caches, a single additive feature over the prior build.

New features

  • KV cache quantization now works with rotated cache layouts, enabling quantized caches for models that use rotation.

Upgrade notes

This is an incremental build with one additive capability; existing flags and model files remain compatible. Rotated-KV models can now be run with a quantized cache, but validate output quality before enabling it in production. No config migration is required.

Analyzed Oct 31, 2026

Upgrading?

Release notes summary

This release adds support for quantizing rotated Key-Value (KV) caches, which is relevant for models utilizing specific cache layouts.

Key Highlights

  • Support for rotated KV cache quantization

Release Notes (excerpt)

<details open> Support rotated kv cache quant (#26180) </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) - Ubuntu x64 (SYCL FP16) **Android:** - Android arm64 (CPU)…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories