LFM2 (Liquid Foundation Models 2)

model
Liquid AI

Efficient foundation models using a hybrid backbone combining gated short convolutions with grouped query attention. 350M to 24B parameters (MoE, 2.3B active). Optimized for edge and on-device deployment.

LFM2-24B achieves 239 tok/s (#3 speed ranking on AA). Also includes vision-language (LFM2.5-VL), speech, and pharmaceutical variants (with Insilico Medicine). AA Intelligence Index: 10 (24B). Apache 2.0.

Paper (arXiv)HuggingFaceOpenRouterArtificial Analysis

Model Details

Architecture MOE
Parameters 24B
Active params 2.3B
AA Intelligence 10

Variants

Name Parameters Notes
LFM2-350M 350M —
LFM2-1.2B 1.2B —
LFM2-2.6B 2.6B —
LFM2-8B-A1B 8.3B —
LFM2-24B-A2B 23.8B —
LFM2.5-350M 350M Released April 2026, trained on 28T tokens with RL, <500MB footprint

Paper

arXiv HTML
open-weighton-devicemoe

Related