Models & Labs
Llama.cpp b9133 Release Enhances Reasoning Models
llama.cpp Releaseshigh confidence
Why it matters
- →Enhances reasoning model capabilities in server and web UI environments.
- →Introduces mechanisms for smoother continuation of generation tasks.
- →Sets the groundwork for future improvements in reasoning models.
Llama.cpp has released its b9133 update, focusing on enhancing support for reasoning models in server and web UI contexts. The update removes blocking assistant prefill and introduces thinking tags to improve the continuation of generation tasks. It also allows reasoning content to persist through reloads by dropping the reasoning guard on the Continue button. This release is a step towards more advanced reasoning capabilities, although channel-based templates remain out of scope for now.
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 b9753 Release Fixes and Enhancements
llama.cpp Releases · June 22, 2026 · Same story
- 2Llama.cpp b9870 Release Fixes Long Reasoning Loops
llama.cpp Releases · July 4, 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.
