Liquid AI Launches LFM3, Six Open-Weight Multimodal Models

Liquid AI released LFM3, a family of six open-weight models spanning text, vision-language, audio, and omni modalities. The family includes LFM3-1.7B (text), LFM3-VL-2B (vision-language with video), LFM3-Audio-2B (multilingual voice and tool use), LFM3-Omni-2B, d1-2B, and d1-Omni-2B, the latter two described as decision models with calibrated probability outputs. The architecture is tri-hybrid, combining 13 short-convolution layers, 5 sliding-window attention layers, and 6 global attention layers across 24 total layers, optimized across 14 processor classes covering GPUs, CPUs, and NPUs, with support for fine-tuning on a single GPU. Liquid AI frames LFM3 as a foundation for building domain-specific multimodal models suited to on-device inference. LFM3-Omni-2B has 2.17B dense parameters, accepts text, image, video, and audio inputs, and produces text and 24 kHz speech output with a 32,768-token context window; it is released under the LFM Open License v1.0. All benchmark scores for LFM3-Omni-2B are vendor-reported by Liquid AI with no published methodology and no third-party reproduction. LFM3-VL-2B scores at least 7.4 points ahead of all models in its size class across 62 vision benchmarks according to Liquid AI, matches the performance of Gemma 4 12B, surpasses Gemini 3.5 Flash Lite, and comes within 2 points of GPT-5.6 Luna. On AutomationBench, LFM3-1.7B scores 41.29 compared to 15.10 for Qwen3.5-9B.