Hugging Face AI Updates: August 21, 2026

1. Liquid AI Publishes LFM2.5-DSpark for Faster Inference

Hugging Face. A Hugging Face blog post from Liquid AI describes LFM2.5-DSpark, which delivers up to 3.2x faster inference. The post details the performance gains and the techniques behind the speedup for the LFM2.5 model family. It targets practitioners looking to cut inference latency and cost on the platform. Source