DeepSeek officially released the stable version of the V4-Flash model and simultaneously launched the API public test. The National Supercomputing Internet Platform promptly launched the model API calling service and model file download. This officially named DeepSeek-V4-Flash-0731 version has undergone extensive post-training work, significantly enhancing its agent capabilities and instruction following, with multiple benchmark test results now comparable to the current strongest closed-source models.
For developers and enterprise users, the access threshold has been minimized. Users no longer need complicated environment configurations. By logging into the official website of the National Supercomputing Internet on the PC end and entering the "Model Services" section through the "Services" menu on the homepage, they can directly access the API calling page with one click. In addition to online calls, users can also download model files and complete the entire development process including fine-tuning, deployment, and adaptation optimization in a single platform using the secure and trustworthy Notebook development environment provided by the platform.
More than 1,700 models gathered, the domestic open-source ecosystem is accelerating
The significance of this launch goes beyond the opening of a single model. The National Supercomputing Internet AI community has currently accumulated more than 1,700 popular open-source large models from both domestic and international sources, covering mainstream series such as DeepSeek, GLM, Qwen, and Kimi, and has fully launched API interfaces for multiple domestic open-source large models, including DeepSeek-V4, MiniMax-M3, and Kimi-K3. This means that developers can compare, call, and switch between multiple models on the same platform, significantly reducing the trial and error costs of AI application development.
More notably, with the official operation of core nodes, the National Supercomputing Internet has become the first national-level super-intelligent integrated computing resource pool with ten thousand cards. From the underlying computing power supply to the upper-layer model services, this national platform is building a complete AI infrastructure chain. When DeepSeek's latest model can be accessed nationwide with a "one-click call," the landing speed and radiation range of domestic large models may be about to experience a brand new turning point.
