China has recently launched its pioneering open-source large model in the marine field, OceanGPT (Cangyuan), in Hangzhou, Zhejiang Province. Primarily developed by Zhejiang University, this model boasts the capability to respond to queries related to marine expertise and interpret multimodal data. By incorporating a domain knowledge-enhanced reasoning mechanism, OceanGPT effectively mitigates hallucinatory errors. In the realm of marine robot manipulation and underwater intelligent systems, OceanGPT can generate natural language robot codes, dramatically enhancing code writing efficiency. Additionally, the model integrates the MCP protocol, enabling multi-robot collaborative operations and facilitating deployment on the edge, thereby augmenting the autonomous operation capabilities of marine equipment. The OceanGPT series comprises multiple versions, including OceanGPT-basic, OceanGPT-omni, and OceanGPT-coder, leveraging a multi-expert hybrid architecture that seamlessly fuses general models with marine-specific capabilities.