{
 "slug": "LFM_(models)",
 "title": "LFM (models)",
 "type": "model",
 "short_desc": "Liquid AI's model line: the proprietary non-Transformer LFM1 generation (2024) and the open-weight LFM2/LFM2.5 edge families built on hybrid convolution-attention blocks.",
 "categories": [
  "Models",
  "Open-weight models"
 ],
 "infobox": {
  "Developer": "Liquid AI",
  "Announced": "LFM1 September 30, 2024; LFM2 July 10, 2025; LFM2.5 January 2026",
  "Architecture": "LFM1: non-Transformer structured operators from dynamical-systems research; LFM2 onward: hybrid gated short convolutions with grouped-query attention",
  "Parameters": "Roughly 230M to 40.3B across generations",
  "License / access": "LFM1 proprietary hosted; LFM2 and LFM2.5 open weights under the Apache-2.0-based LFM Open License v1.0",
  "Paper/report": "LFM2 technical report, arXiv:2511.23404"
 },
 "infobox_links": {
  "Developer": [
   "Liquid_AI"
  ],
  "Architecture": [
   "Hybrid_architecture_(LLM)"
  ]
 },
 "aliases": [
  "Liquid Foundation Models",
  "LFM2"
 ],
 "family": "",
 "org": "Liquid AI",
 "registry_status": "complete",
 "words": 562,
 "references": 5,
 "outbound": [
  "Falcon_(models)",
  "Gemma",
  "Hugging_Face",
  "Hybrid_architecture_(LLM)",
  "Jamba",
  "Knowledge_distillation",
  "Large_language_model",
  "Liquid_AI",
  "Mixture_of_experts",
  "Phi_(model_family)",
  "Qwen3",
  "State-space_model",
  "Transformer_(architecture)"
 ],
 "inbound": [
  "Liquid_AI",
  "Stated_missions_of_AI_labs"
 ],
 "url": "wiki/LFM_(models).html",
 "built": "2026-07-24"
}