Liquid AI's smallest model yet LFM2.5-230M beats models 4X its size at data extraction, can run 'anywhere'

Liquid AI released a small artificial intelligence language model designed to run on devices like smartphones, laptops, and robots while performing data extraction tasks. The model achieves performance comparable to models significantly larger than itself, outperforming some models four times its size on specific benchmarks. This matters because enterprises currently use expensive cloud-based AI services for routine data formatting and extraction, whereas this smaller model can operate locally on hardware, reducing costs and latency. The release reflects a broader shift in the AI industry toward building efficient models optimized for edge devices rather than pursuing ever-larger parameter counts.