Models & Labs
Llama.cpp b10704 Release Optimizes CUDA Path
llama.cpp Releaseshigh confidence
Why it matters
- →Enhances CUDA performance for models using multiple experts.
- →Increases prompt processing speed, benefiting high-performance applications.
- →Maintains broad platform support, ensuring accessibility for developers.
Llama.cpp's b10704 release brings an important optimization for CUDA users, specifically targeting the fast mm_ids_helper path. This update allows configurations with n_expert_used = 10 to access the fast path, resulting in improved prompt processing speeds on an RTX PRO 6000. The optimization does not affect token generation but enhances performance for models using multiple experts. The release maintains support across various platforms, including macOS, Linux, Windows, and more.
Read originalThe story around this
Earlier coverage that leads up to this article, and what followed. Lines connect each piece to the closest one after it, converging here.
How we got here
- 1Llama.cpp b9866 Update Enhances CUDA Performance
llama.cpp Releases · July 4, 2026 · Same story
What happened next
- 2Llama.cpp b10751 Release Enhances CUDA Efficiency
llama.cpp Releases · September 2, 2026 · Same story
More from llama.cpp Releases
Coding Toolscoding
llama.cpp b12022 adds ROCm 10 and Snapdragon support
This release quietly cements llama.cpp as the universal inference runtime by finally bringing ROCm 10.0 to Linux and Windows alongside CUDA 13.4, effectively closing the hardware gap for AMD users who were previously left behind. The inclusion of native Snapdragon support on Linux marks a significant step toward running local LLMs on ARM-based mobile chips without emulation overhead. While KleidiAI on Apple Silicon is disabled in this build, the broader expansion to diverse accelerators ensures that developers aren't locked into NVIDIA ecosystems anymore.
llama.cpp Releases
Coding Toolscoding
llama.cpp b12023 fixes shared prefix bug
This release addresses a specific inference artifact where decision prefixes were incorrectly bleeding into output embeddings. By capping shared prefixes at the first output position, llama.cpp ensures that option tokens remain distinct for parent and child tasks. This is a targeted fix for multi-step reasoning or voting workflows rather than a performance boost. It stabilizes complex generation patterns without changing the underlying model architecture.
