Models & Labs
HMX Flash Attention Update Released
llama.cpp Releasesmedium confidence
Why it matters
- →Enhancements in flash attention can lead to better performance in AI models. • The update addresses critical correctness issues, ensuring reliable outputs. • Optimizations may reduce computational overhead, benefiting developers working with large datasets.
The HMX Flash Attention update introduces multiple enhancements, including HMX-accelerated flash attention for prefill and optimizations for multi-threading and memory management. Key changes involve replacing inline assembly with Q6 intrinsics for better compiler visibility, refining cost model coefficients based on profiling data, and fixing correctness issues related to prefill and softmax operations. This update aims to improve performance and maintainability of the HMX Flash Attention implementation, making it more efficient for users.
Read originalMore 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.
