Models & Labs
llama.cpp b9974 Release Fixes CUDA Memory Crash
llama.cpp Releaseshigh confidence
Why it matters
- →Prevents crashes on CUDA devices with no free memory, improving stability.
- →Ensures the fit algorithm doesn't attempt to use unsuitable devices.
- →Enhances reliability for developers using CUDA-enabled builds.
The b9974 release of llama.cpp resolves a significant issue for CUDA users by preventing crashes when querying memory on devices with no available memory. Previously, the call to cudaMemGetInfo() could cause a fatal crash if the device was out of memory. The update now assigns zero total/free memory to such devices, preventing the fit algorithm from attempting to use them and avoiding crashes. This update enhances stability for CUDA-enabled builds, particularly when users specify '-dev none'.
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 b9047 release focuses on device memory handling
llama.cpp Releases · May 7, 2026 · Same story
- 2llama.cpp b9274 Release Fixes VRAM Leak
llama.cpp Releases · May 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.
