Liquid AI's LFM3 family of small multimodal foundation models for on-device, edge and cloud use.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
LFM3 is a generation of Liquid Foundation Models from Liquid AI, announced on 27 October 2026 under the title "Every Modality for Specialized Intelligence". It targets on-device and edge deployment, with models around 2B parameters or smaller. According to YouTube coverage, the family covers text, vision, audio, omni and decision variants, and was published in many quantized formats, including MLX.
Creators report that the 1.7B model outperforms the previous 2.6B model, fits in roughly 1 GB at 4-bit, supports a 256K context window, and handles text, video and reasoning. These figures come from third-party summaries and have not been checked against Liquid AI's own benchmarks.
Liquid AI states that its open-weight models are free to download, run and fine-tune. Commercial use is free for companies under $10M in annual revenue, and larger companies need a paid enterprise license. The models are released under the LFM Open License. The weights run on runtimes such as llama.cpp, MLX, ONNX, Ollama, SGLang and vLLM. A hosted API is also listed, with LFM3-1.7B priced at $0.08 per million input tokens and $0.20 per million output tokens.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
New generation of Liquid foundation models for on-device use. The 1.7B version outperforms the previous 2.6B, fits in about 1 GB at 4-bit, has a 256K context, and handles video, text and reasoning.
Family of six ~2B models (text, vision, audio, omni, decision) released on Hugging Face in many quantized variants including MLX. Text model outperforms larger competitors.
Closest neighbours by meaning, not just by category.
Model · Liquid AI
Liquid AI's 24B-parameter mixture-of-experts LFM with 2B active parameters, built for fast on-device agents.
Model · OpenBMB
Small open-weights multimodal model from OpenBMB for image and video understanding on phones and edge devices.
Model · Alibaba
Small 0.8B-parameter open-weights language model in Alibaba's Qwen 3.5 family, aimed at lightweight and on-device use.
Model · Xiaomi
Xiaomi's MiMo family of open-weight language, multimodal, audio and speech models, with a hosted API and web demo.
Model
Compact 1.7B open-weights text LLM for on-device use, with adjustable thinking levels and English/Chinese support.
Model · Google
Google DeepMind's family of open-weights models built from Gemini technology, from phone-sized to laptop-class variants.
