Tencent Hybrid releases multimodal understanding model Hybrid Large-Vision.

On August 12th, Tencent released the multimodal understanding model Hybrid Large-Vision. It adopts the MoE architecture with 52 billion activation parameters, and supports input of images, videos, and 3D space at any resolution, focusing on improving the ability to understand multilingual scenarios.