Recently, DeepSeek's entire series of large models, including DeepSeek-R1, V3, and Coder, have officially connected to the National Supercomputing Internet Platform (www.scnet.cn). Supported by domestic super-intelligent integrated computing power and a national unified computing service system, China's AI large model industry has taken a crucial step forward.
The National Supercomputing Internet Platform has currently launched versions of the DeepSeek-R1 model, including 1.5B, 7B, 8B, and 14B. Users can directly experience online inference services through QR code scanning or web access, without the need for local deployment. For enterprise-level needs, the platform supports the introduction of proprietary data for customized private model development, with larger parameter versions like 32B and 70B set to be released soon. Also launched are vertical models such as DeepSeek-V3, Coder series, and Math series (7B), covering diverse scenarios like code generation and mathematical reasoning.
Compared to overseas large models, the DeepSeek series significantly reduces computing resource consumption through algorithm optimization. Combined with the National Supercomputing Platform's capabilities for "affordable and easy-to-use" computing resource scheduling, this leads to a substantial reduction in invocation costs. For example, the DeepSeek-R1-Distill-Qwen-7B features a WebUI dialogue interface that supports a no-threshold experience, while enterprise users can quickly integrate into production environments via API interfaces, creating a complete link from model invocation to private deployment.
Analysts point out that the large-scale implementation of DeepSeek models on the supercomputing platform may change the current situation in the domestic AI industry, which relies on overseas closed-source models. As more developers train and optimize localized models based on domestic computing power, China is expected to achieve a substantial "leapfrog" in the application layer of AI technology. This wave of exploration driven by "mysterious Eastern power" is injecting new variables into the global AI competition.
