Models & Labs
Llama.cpp adds GLM-5.2 speculative decoding support
llama.cpp Releaseshigh confidence
Why it matters
- →Enhances model performance with speculative decoding for GLM-5.2.
- →Provides developers with flexible export options for model configurations.
- →Optimizes tensor loading and context management for improved efficiency.
Llama.cpp has released an update adding speculative decoding support for the GLM-5.2 model, specifically targeting the GLM_DSA architecture. This enhancement includes NextN/MTP features, which improve tensor loading and context management. Developers can now export models with or without the MTP feature, offering greater flexibility. This update is significant for optimizing model performance and adaptability, particularly for those using the GLM-5.2 framework.
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 adds MiniMax model support
llama.cpp Releases · August 16, 2026 · Same story
- 2llama.cpp v0.3.0 Release
llama.cpp Releases · August 26, 2026 · Same story
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.
