Models & Labs
Llama.cpp b9689 Release Adds Metal Backend Support
llama.cpp Releaseshigh confidence
Why it matters
- →Expands Metal backend capabilities to support more tensor types.
- →Enhances AI model performance on Apple Silicon devices.
- →Increases versatility of llama.cpp across different platforms.
Llama.cpp has released version b9689, which introduces significant updates to its Metal backend. The concat operator now supports f16 and bf16 tensor types, in addition to the existing f32 and i32. This enhancement is achieved by templating the kernel_concat on type T and adding type-specific pipeline getters. The update is particularly beneficial for macOS and iOS developers, as it improves AI model performance on Apple Silicon devices. This release marks a step forward in making llama.cpp more versatile across different platforms and data types.
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 b10159 release enhances Metal backend
llama.cpp Releases · July 29, 2026 · Same story
- 2llama.cpp b10414 Release Adds TQ2_0 Support
llama.cpp Releases · August 14, 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.
