archive
· today in ai · 2026-08-13
Liquid AI ships a 3B vision model for phones
Archive item — written before sources were shown.
LFM2.5-VL-3B reads screens, grounds objects, and calls tools from a 3GB footprint, matching larger 4.7B rivals while running on a single device.
Liquid AI released open weights for LFM2.5-VL-3B on August 12, a 3.1B-parameter vision-language model built to run on phones, laptops, and single GPUs. It reads screens and documents, grounds objects to coordinates, and calls tools from text or image input, scoring 69.4 average across 28 vision benchmarks, on par with other ~4.7B models, while fitting in roughly 3GB of memory.
sources
- 01LFM2.5-VL-3B: A Better and Faster Vision-Language Model for the Edgeliquid.ai · primary
Rami Steitieh