AntGroup Open Sources Ling-3.0-flash-VL: The 124B-Parameter Native Multimodal Large Model Makes Its Debut
The AntGroup Open Source team released the first native multimodal large model of the BaiLing series, Ling-3.0-flash-VL. It is based on the MoE architecture of Ling-3.0-flash, with a total parameter count of 124B and an activated parameter count of 5.5B. It natively supports image, text, and video input, with a context window of 256K tokens. The team focused on improving reliability and efficiency in real tasks, and made efforts to address the issue where visual capabilities might weaken text intelligence.