TMTPOST —  The National Supercomputing Internet Platform has officially launched multiple DeepSeek AI models, including the DeepSeek-R1 model. The platform offers one-click inference services for the smaller versions of DeepSeek-R1, eliminating the need for local downloads. Additionally, users can introduce proprietary data to customize and train models based on private deployment needs. Beyond DeepSeek-R1, the platform has also integrated a range of other models, including DeepSeek-V3, DeepSeek-V2.5 series, DeepSeek-Coder series, DeepSeek-Math series (7B), and DeepSeek-V2 series (Lite).