Ant Group · Multimodal MoE LLM

Ling-3.2-flash (Ling model series)

Ling is a general-purpose large language model series developed and open-sourced by Ant Group, built on a Mixture of Experts architecture. The page documents the series and highlights Ling-3.2-flash, described as the latest native multimodal model, with 560B total parameters, 25B activated parameters and a context window of up to 1M. It accepts image, text and video input and focuses on complex visual understanding, multimodal long-horizon agents, 3D understanding and interaction, and medical multimodal tasks. The page also lists earlier models: Ling-3.1-flash (560B/25B, 1M context, aimed at office work and long-horizon software development), Ling-3.0-flash-VL (124B/5.5B, 256K), Ling-3.0-flash (124B/5.1B, 256K, peak speed stated as up to 1000 tokens/s) and Ling-3.0-tiny (7.9B/1.3B, for on-device use). No benchmark results, licence terms or release dates appear in the text provided.