Models & Labs
llama.cpp b9935 release adds VISION RoPE support
llama.cpp Releaseshigh confidence
Why it matters
- →VISION RoPE support enhances data handling capabilities in llama.cpp.
- →Decoupling DMA copy size from row stride allows for more flexible data processing.
- →Support for non-contiguous destinations improves versatility in data management.
The b9935 release of llama.cpp brings VISION RoPE support, enhancing its ability to manage strided half-dim views in all modes. This update decouples the source DMA copy size from the row stride, allowing for more flexible data handling. It also supports non-contiguous destinations for RoPE, improving data processing capabilities. While no new models are introduced, these changes enhance the existing framework, offering developers more versatility in handling complex data structures.
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 Adds Vision Support with MiniMax-M3
llama.cpp Releases · July 27, 2026 · Same story
- 2llama.cpp b10568 Release Expands Platform Support
llama.cpp Releases · August 22, 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.
