Models & Labs
Llama.cpp b9873 Release Fixes K/V Rotation Bug
llama.cpp Releaseshigh confidence
Why it matters
- →Fixes a critical bug that could cause system aborts during tensor operations.
- →Enhances the stability and reliability of llama.cpp for developers.
- →Improves handling of complex AI model operations, making the platform more robust.
Llama.cpp has released version b9873, which fixes a bug related to K/V rotation input handling when the buffer is unallocated. The update prevents the system from aborting by adding a guard for k_rot/v_rot inputs, similar to the existing checks for kq_mask inputs. This change is particularly relevant for processes like DFlash speculative decoding's KV-injection pass. The update enhances the stability of llama.cpp, ensuring smoother operations for developers using the platform.
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 b9871 Release Fixes CPU Concat Issue
llama.cpp Releases · July 6, 2026 · Same story
- 2llama.cpp b10593 Release Fixes and Enhancements
llama.cpp Releases · August 24, 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.
