Models & Labs
Llama.cpp b9075 Release Enhances CUDA Snake Activation
llama.cpp Releaseshigh confidence
Why it matters
- →The fusion of snake activation into a single kernel improves computational efficiency.
- →Supports multiple data types, enhancing versatility for developers.
- →Optimizes performance for audio decoders, a critical application area.
Llama.cpp's b9075 release focuses on optimizing CUDA performance by fusing the snake activation function into a single kernel. This change targets audio decoders such as BigVGAN and Vocos, which previously used a more complex operation sequence. The update supports F32, F16, and BF16 data types, enhancing efficiency and performance. This release highlights llama.cpp's ongoing efforts to improve CUDA functionality, offering developers a more streamlined and effective tool for handling intricate activation functions.
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.
What happened next
- 1llama.cpp b9112 Release Fixes CUDA Limitations
llama.cpp Releases · May 12, 2026 · Same story
- 2Llama.cpp b10099 Release Enhances CUDA Quantization
llama.cpp Releases · July 25, 2026 · Same story
Follow this story
Open the full 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.
