Today for AI
HOT RADAR
llmHEAT 33.9°

AI CURATED HOT STORY · 10/28/2026

Liquid AI Launches LFM3 Open-Source Models, Outperforming Predecessors

2 reports archived2 independent sourcesupdated 10/28/2026, 01:51:06
Synthesis & Latest Updates
2 Sources Cross-Validated

Liquid AI has launched the LFM3 open-weight model family, marking a significant breakthrough in edge AI. The series features an innovative "tri-hybrid" architecture combining short convolutions, sliding window attention, and global attention to maximize efficiency. Notably, the 1.7B parameter text model outperforms its 2.6B predecessor and larger competitors like Qwen3.5-9B in automation benchmarks, scoring 41.29 versus 15.10 on AutomationBench.

LFM3 is a comprehensive ecosystem comprising six instances covering text, vision, audio, and full modality capabilities. Built on 22T tokens of pre-training data, it unifies these modalities into a single shared backbone, making it the first 2B-class multimodal family to integrate decision-making with perception. All weights are fully open-sourced on Hugging Face, supporting local fine-tuning and execution on any device, thereby lowering barriers for building specialized agents.

This release reshapes the landscape for edge computing and privacy-focused applications. By demonstrating that small models can achieve high performance through architectural innovation, LFM3 facilitates the shift from cloud-centric to device-centric AI. Developers can now cost-effectively customize high-performance vertical agents, accelerating AI adoption in consumer electronics and IoT sectors.

LATEST/LFM3 weights are now available on Hugging Face for immediate local deployment and fine-tuning. Early community feedback highlights its utility in low-latency edge agents, likely sparking a new wave of competition in small model optimization.

Hourly heat curve

4 fully observed hours
Now
33.9
Peak
35.502:00
24h change
–

The line compares only sources observed throughout; gap hours are interpolated to keep the trend continuous.

Coverage timeline

Total 2 reports · Latest first
  1. Rohan Paul (@rohanpaul_ai)T2·85 pts

    Liquid AI Launches Open-Weight LFM3 Family with Tri-Hybrid Architecture

    • LFM3 uses a tri-hybrid architecture (short convolutions + sliding-window attention + global attention) to boost small-model efficiency.
    • The 1.7B text model scores 41.29 on AutomationBench, significantly outperforming Qwen3.5-9B's 15.10.
    • The family includes six model instances covering text, vision-language, audio, and omni-modalities, optimized for on-device execution.
  2. Hacker News AIT2·78 pts

    Liquid AI Releases Open-Weight Multimodal LFM3 Models for Local Fine-Tuning

    • LFM3 is the first 2B-parameter multimodal family unifying text, vision, audio, and decision models onto a single shared backbone.
    • Built on a new hybrid architecture with 22T tokens pre-training, it offers greater expressiveness than LFM2 while maintaining high efficiency.
    • Fully open-weight models support running on any device and local fine-tuning, ideal for building specialized intelligent agents.