Liquid AI Releases 230M Ultra-Small Model

dl_weekly · x · 2026-07-09

Liquid AI has released its smallest open-weight model to date, LFM2.5-230M. The model achieves an inference speed of 213 tokens per second on the Samsung Galaxy S25 Ultra and outperforms larger models in benchmarks for tool use and data extraction.

Original post →

More from Infra

Infra channel →