b10068
ggml-org/llama.cppb10068Jul 18, 2026by github-actions[bot]same day as b10067
What changed since b10067
Fixes DeepSeek-V4 quantization and DFlash K/V cache rotation.
Improvements
- Improves DFlash model compatibility
- Rotates injected K/V cache when using K/V quantization
Bug fixes
- Fixes DeepSeek-V4 routing table quantization
- Fixes DFlash K/V cache handling with quantization
Upgrade notes
No configuration changes are required. This update fixes specific issues with DeepSeek-V4 and DFlash models.
Analyzed Oct 16, 2026
Upgrading?
Release notes summary
DFlash model update for proper handling of injected K/V cache.
Key Highlights
- Rotate injected K/V cache when using K/V quantization
- Fixes issue #25823
- Improves DFlash model compatibility
Release Notes (excerpt)
<details open> model: rotate injected K/V cache for DFlash (#25823) * dflash: rotate injected K/V cache when using K/V quantization * Update src/models/dflash.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * clearer format * remove trailing whitespace --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> </details> **Website:** - <https://llama.app> **macOS/iOS:** - macOS Apple Silicon (arm64) - macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED…
Published by ggml-org on GitHub.
Read the full release notesRelated Repositories
:electron: Build cross-platform desktop apps with JavaScript, HTML, and CSS
Godot Engine – Multi-platform 2D and 3D game engine
ClickHouse® is a real-time analytics database management system
Self-hosted game stream host for Moonlight.
DuckDB is a high-performance analytical database system