Liquid AI Releases LFM3: Six Open-Weight Multimodal Models on Shared Hybrid Backbone
Liquid AI released the LFM3 series, comprising six open-weight 2B-class models for text, vision, audio, and decision tasks built on a shared tri-hybrid backbone (gated short convolutions + sliding-window attention + global attention). Optimized via hardware-in-the-loop search, LFM3-1.7B outperforms Qwen3.5-9B on agentic benchmarks like BFCL V4 and WebArena but trails significantly in long-context reasoning (RULER) and complex knowledge tasks (GPQA Diamond).
Full Report
▲ Marktechpost AI · 2026-10-27 21
▲ Liquid AI just released LFM3, 6 open-weight 2B-class models that put text, vision, audio, omni and decision tasks on 1 shared backbone.
The key idea is a tri-hybrid architecture. Its 24 layers mix 13 gated short convolutions, 5 sliding-window attention layers and 6 global attention layers. Liquid tuned the mix with hardware-in-the-loop search across 14 device and backend combinations.
On Liquid's own evals, LFM3-1.7B averages 49.37 across 29 benchmarks, vs 47.24 for Qwen3.5-9B. It scores 71.39 on BFCL V4 vs 62.83, and 38.42 on WebArena-Verified vs 28.94. It is a trade-off though: GPQA Diamond is 63.37 vs 80.40, and RULER 128K is 65.56 vs 87.11.
LFM3-1.7B has 1.73B parameters, a 256K context window and takes about 1 GB at 4-bit. Weights ship under the LFM Open License v1.0, free for commercial use below $10M annual revenue......
Full analysis:
