launchLiquidAI2026-10-28
Liquid AI releases LFM3 model family, six models based on a 1.7B base with 256K context
Liquid AI released the LFM3 model family on October 27 local time with six first models, all built on LFM3-1.7B-Base, using a new tri-mixture layer architecture and pretrained on 22 trillion tokens. The family covers text, vision, decision, audio and omni modalities with LFM-1.7B, LFM3-VL-2B, d1-2B, LFM3-Audio, LFM3-Omni-2B and d1-Omni-2B, where the text model offers a 256K token context and LFM3-Audio carries no parameter size in its name. Liquid AI says LFM-1.7B performs better than the previous LFM2.5-2.6B while using about 35 percent fewer parameters. The report does not provide pricing, a weight license name, download channels or actual availability.
Liquid AILFM3模型家族三混合层架构256K 上下文