Hunyuan-Large-Vision

Multimodal understanding model built on MoE architecture (389B total, 52B active). Handles images, videos, and 3D content. Ranked first among Chinese image AI models on LMArena Vision Leaderboard.