LFM2.5-24B-A2B: Fast, Private On-Device AI Model Released
Liquid AI has released LFM2.5-24B-A2B, a 24B-parameter reasoning model designed for fast, private agents on-device. Despite having only 2B active parameters, it competes with larger models like Qwen3.6-35B-A3B on several benchmarks, achieving significant performance gains on tasks like AutomationBench. The model features an expanded vocabulary and context window, along with scaled-up pretraining and reinforcement learning. It is optimized for edge and GPU inference, running up to 1.7× faster than comparable models on-device and reaching over 21K output tokens per second on a single NVIDIA B300 GPU. The base and post-trained models are open-weight and available on Hugging Face and the company’s Playground.
Read the full article
Continue reading at Hacker News →This is an AI-generated summary. Read the original for the full story.