{
  "id": 16845429,
  "title": "Liquid AI's LFM3: Six 2B Open Models, One Shared Brain — and a 1GB Phone-Sized Agent",
  "url": "https://urgent.news/2026/10/28/liquid-ais-lfm3-six-2b-open-models-one-shared-brain-and-a-1gb-phone",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-28T04:17:45.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/paullovelace/liquid-ais-lfm3-six-2b-open-models-one-shared-brain-and-a-1gb-phone-sized-agent-40co"
  },
  "original_language": "en",
  "account": "On October 27, 2026, Liquid AI unveiled LFM3, a collection of six open-weight models with around 2 billion parameters each, all branching from a shared backbone. The models are designed to run on devices with limited resources, such as phones, laptops, and Raspberry Pis. The text model, at 4-bit precision, requires only 1 GB of memory, making it small enough to live on the device it serves. These are the key points of the story: Liquid AI announced LFM3 on October 27, 2026, the models branch from a shared backbone, are designed for low-resource devices, and are free under the LFM Open License v1.0 for companies with annual revenue under $10 million.",
  "summary": "Liquid AI announced LFM3 on October 27, 2026, and the pitch is the opposite of the usual bigger-is-better launch: six open-weight models, all around 2 billion parameters, all branching from one shared backbone, built to run where the big models cannot — on a phone, a laptop, a Raspberry Pi. The text model, at 4-bit precision, takes about 1 GB of memory. That is the whole story in one number: an…",
  "key_points": [
    "Liquid AI unveiled LFM3 on October 27, 2026",
    "Models branch from shared backbone, designed for low-resource devices",
    "Free under LFM Open License v1.0 for companies with annual revenue under $10 million"
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}