Skip to content
NewA calmer, faster ZeroHour
StoryModel releaseinfo

Liquid AI announces open-weight LFM3 models near 2B parameters

3 sources · 3 articlesFirst reported Updated ()1
What's new

First merged summary.

Merged summary · x-ai/grok-4.7 · rewritten as coverage arrivesSummaries by Vani AI

Liquid AI has announced LFM3, open-weight models near 2B parameters, with LFM3-1.7B offering a 256K context.

Liquid AI announced LFM3, a family of Liquid Foundation Models around 2B parameters that share one backbone for text, vision, audio, omni, and decision models. Hacker News reports the first release is open-weight and available on Hugging Face, the Liquid AI Playground, and the Liquid AI API. LFM3-1.7B-Base was pre-trained on 22T tokens; a final 3.5T-token mid-training stage extends context to 256K. MarkTechPost’s headline says six open-weight 2B-class multimodal models, while Hacker News does not state that count. The Hugging Face blog and MarkTechPost page bodies do not add license, benchmark, or model-card details.

Key facts
  • LFM3-1.7B-Base was pre-trained on 22 trillion tokens.
  • Context reaches 256K after a 3.5T-token mid-training stage.
  • Architecture: 13 convolution, five sliding-window, six global-attention layers.
  • Liquid AI says it is about 25% faster per parameter than LFM2-1.2B at 8K.
  • Post-trained for Hermes Agent, Pi, and OpenClaw harnesses.
  • Hardware search covered 14 device and backend combinations.
Explain this AI news with Vani AIWhat was announced, what is new and why it matters, in a new chat

Timeline

dated events in the reporting · oldest first
  1. Oct 27, 2026Liquid AI announces the LFM3 model family. · Hacker News

What each outlet adds

beyond the shared facts
  • Hacker NewsArchitecture, token counts, context, speed claim, and availability.
  • MarkTechPostHeadline states six open-weight 2B-class multimodal models.
Vendors
Liquid AI
Organizations
AMDAmbarellaLiquidAIHugging Face
AI models
LFM3LFM3-1.7BLFM3-1.7B-Base

Coverage

oldest first · one row per article · UTC
  1. · 3d ago Hacker News · securityFirst report

    Liquid AI releases open-weight multimodal LFM3 models around 2B

    LFM3-1.7B-Base was pre-trained on 22T tokens, with a 256K context after mid-training.

  2. · 3d ago Hugging Face Blog

    Hugging Face blog post covers building models with LFM3

    No article body was available, so LFM3's capabilities, license, and availability are not stated.

  3. · 3d ago MarkTechPost

    Liquid AI releases LFM3, six open-weight 2B models

    The available text is only the headline; licenses, benchmarks, and release details are not included.