Models & Labs
Llama.cpp Adds Adreno Support with New Kernels
llama.cpp Releaseshigh confidence
Why it matters
- →Expands llama.cpp's compatibility with Qualcomm's Adreno GPUs.
- →Enhances performance for AI tasks on Adreno-equipped devices.
- →Broadens hardware support, making llama.cpp more versatile.
Llama.cpp has released version b9603, which includes new OpenCL kernels for Adreno GPUs, specifically targeting q5_0 and q5_1 gemm and gemv operations. This update, co-authored by Li He from Qualcomm, aims to improve performance on devices with Adreno graphics. The release does not feature new models but focuses on expanding hardware support, particularly for Qualcomm's Adreno GPUs. This development allows for better utilization of llama.cpp on a broader range of devices, enhancing its applicability in AI tasks.
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
- 1b9070 release adds Q4_0 MoE GEMM for Adreno
llama.cpp Releases · May 8, 2026 · Same story
- 2Llama.cpp b9127 Release Adds Adreno xmem GEMM
llama.cpp Releases · May 13, 2026 · Same story
Follow this story
Open the full story →llama.cpp b9571 Release Expands Platform Support
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.
