DeepSeek V4 Pro and Harness Launched on National Supercomputing Internet

en.Wedoany.com Reported - Dimension News, the National Supercomputing Internet announced on August 14 the launch of the official version of DeepSeek V4 Pro (DeepSeek-V4-Pro-0813) and the agent framework DeepSeek Harness. As the nation's first 100,000-card-scale super-intelligent integrated computing resource pool, the National Supercomputing Internet can provide research institutions, tech enterprises, and developers with full-lifecycle computing support for large models, from training and fine-tuning to evaluation and deployment.

DeepSeek-V4-Pro-0813 places particular emphasis on enhanced Agent capabilities, with notably significant improvements in production environment performance. On August 13, the DeepSeek Harness developer preview (v0.1) was opened for testing to developers worldwide and open-sourced simultaneously under the MIT license. The Supercomputing Internet AI community has made the full source code of Harness available online, allowing developers to pull project files with one click from the platform's code repository for compilation, debugging, secondary iteration, and project deployment.

Currently, the Supercomputing Internet AI community has aggregated over 1,700 open-source large models, enabling users to complete the entire AI development workflow—including model deployment, fine-tuning training, and adaptation optimization—in a one-stop manner.