Ant group

Ant Ling launches a multimodal model for million-token visual agents

Ling-3.2-flash accepts text, images and video, activates 25 billion of 560 billion parameters and is aimed at long-running visual tasks.