Search
Ant Group's Bailing large model team has recently made available the open-source version of the unified multi-modal large model, Ming-lite-omni. Built upon the foundation of Ling-lite, this model adopts a Mixture-of-Experts (MoE) architecture and features a total of 22 billion parameters. In various benchmarks, its performance rivals that of leading models in the 10 billion parameter range, positioning it as the first open-source model with modal support nearing the capabilities of GPT-4. The team emphasizes their commitment to enhancing Ming-lite-omni's full-modal task performance and complex reasoning abilities, and they plan to train a larger variant, Ming-plus-omni. Furthermore, they are in the process of developing the Max version of Ling.
-
C114 Communication Network -
Communication Home
