Liquid AI Launches LFM2.5-VL-3B: A Lightweight On-Device Vision-Language Model

JosephJacks_ · x · 2026-08-13

Liquid AI has released LFM2.5-VL-3B, a lightweight vision-language model built for mobile, web, and desktop environments. It is capable of reading screens, documents, and the physical world.

Built on the LFM2.5-2.6B base with a SigLIP2 400M vision encoder and pre-trained on 34T tokens, it achieves comparable or better scores than models up to 2.6x its size. It scores 80.7 on ScreenSpot-v2 and 73.1 on RealWorldQA, while supporting tool calling from both text and image inputs.

Related event: LiquidAI Releases LFM2.5-VL-3B Edge Multimodal Model(7 posts)→

Original post →

More from Models

Models channel →