Liquid AI Releases LFM2.5-VL-450M: a 450M-Parameter Vision-Language Model with Bounding Box Prediction, Multilingual Support, and Sub-250ms Edge Inference

Liquid AI released LFM2.5-VL-450M, a 450M-parameter vision-language model tuned for edge devices.

The update adds bounding box prediction, stronger instruction following, broader multilingual support, and text-only function calling.

It is designed for sub-250ms inference on hardware such as Jetson Orin, AMD Ryzen AI Max+ 395, and Snapdragon 8 Elite devices.

The added grounding and structured-output features expand on-device use cases beyond captioning to detection and workflow automation.