LFM2-VL is a new series of open-weight vision-language models from Liquid AI. Designed for on-device deployment, they offer up to 2x faster inference on GPU and come in 450M and 1.6B parameter sizes.

LFM2 by Liquid AI is a new class of open foundation models designed for on-device speed and efficiency. Its hybrid architecture delivers 2x faster CPU performance than Qwen3 and SOTA results in a tiny footprint.