Models & Labs
Llama.cpp Update Fixes Reasoning Budget Handling
llama.cpp Releaseshigh confidence
Why it matters
- →Fixes ensure caller-supplied values are respected, enhancing AI control.
- →Improves precision in AI reasoning processes for chat completions.
- →Validated with unit tests, ensuring reliability and accuracy.
Llama.cpp has released an update that fixes a bug related to reasoning budgets in chat completions. The issue involved the system ignoring per-request values for reasoning_budget_tokens and reasoning_budget_message, defaulting to server-level settings instead. This update ensures that the caller's intended values are used, enhancing the precision of AI reasoning processes. The fix is verified with unit tests, improving the flexibility and accuracy of chat completions.
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.
How we got here
- 1Llama.cpp b9133 Release Enhances Reasoning Models
llama.cpp Releases · May 14, 2026 · Same story
- 2Llama.cpp introduces real-time reasoning control
llama.cpp Releases · June 2, 2026 · Same story
Follow this story
Open the full story →Llama.cpp b9870 Release Fixes Long Reasoning Loops
More from llama.cpp Releases
Coding Toolscoding
llama.cpp b11183 splits Metal FA kernels
This release quietly refactors how llama.cpp handles Flash Attention on Apple Silicon by splitting kernels into per-dtype libraries. It’s a structural optimization that likely reduces memory overhead and improves compilation times for Metal users, though the immediate performance gains are subtle compared to algorithmic leaps. The build matrix remains massive, adding ROCm 10.0 and CUDA 13.4 support while disabling KleidiAI on Apple Silicon for now. This is infrastructure maintenance rather than a feature breakthrough, but it keeps the runtime robust across the expanding landscape of hardware backends.
llama.cpp Releases
Coding Toolscoding
llama.cpp b11184 expands Metal support to wider blocks
This release tackles a specific performance bottleneck on Apple Silicon by extending Metal FWHT kernels to handle block widths up to 8192. Previously limited to 512, these wider operations now use threadgroup memory instead of registers, enabling faster inference for larger context windows or model architectures that rely on Hadamard transforms. The update also ships binaries for CUDA 13 and ROCm 10.0, keeping the toolkit aligned with the latest NVIDIA and AMD driver ecosystems. It’s a quiet but necessary optimization that improves throughput on M-series chips without changing the user experience.
