Models & Labs
Hugging Face Releases LFM2.5 QAD Checkpoints
Hugging Face Bloghigh confidence
Why it matters
- →Quantization-Aware Distillation significantly improves model performance while maintaining efficiency.
- →The release enables advanced AI capabilities on edge devices, broadening accessibility.
- →The models recover a high percentage of accuracy lost to quantization, enhancing practical usability.
Hugging Face has released new LFM2.5 Q4_0 checkpoints trained with Quantization-Aware Distillation (QAD), which significantly improve performance while retaining low memory and high throughput. These models recover 97% of the accuracy lost to quantization, matching or exceeding the quality of higher precision models with increased decode throughput. The release includes models ranging from 230M to 2.6B parameters, optimized for various hardware platforms. This development enhances the deployment of efficient AI models on edge devices, broadening access to advanced AI capabilities.
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.
More from Hugging Face Blog
More in Models & Labs
Models & Labsmodels
llama.cpp b12029 adds CUDA 13 and Snapdragon support
This release quietly cements llama.cpp as the universal inference runtime by finally supporting NVIDIA's latest CUDA 13 stack alongside AMD's ROCm 10.0. The inclusion of native builds for Linux arm64 with Snapdragon chips, covering CPU, Adreno GPU, and Hexagon NPU, signals a serious push into mobile edge deployment. While Apple Silicon KleidiAI is disabled in this specific build, the broader hardware coverage means developers no longer need to compile from source for most modern accelerators. It’s less about new features and more about removing friction for anyone running local models on bleeding-edge or diverse hardware.
llama.cpp Releases
© The Verge AIModels & Labsother
