🌙 Moondream 3.1

A lightweight vision-language MoE model (9B total, 2B active) specializing in visual reasoning, object detection, pointing, and visual querying.

Model card · Blog

Task
Examples