Ant launches Ling-3.2-flash multimodal model with million-token context

Ant Group released the Ling-3.2-flash native multimodal model, which accepts text, images, and video, features 560 billion total parameters with 25 billion activated per inference step, and supports a one-million-token context window. According to the company, the model is designed for complex visual understanding, multistep agent tasks, 3D interaction, and medical applications. It is accessible via an online chat interface and APIs compatible with OpenAI and Anthropic protocols, offering free access for its first two weeks. Open article →