【Frontiers of HaiChuan Technology】Sinopec Takes the Lead in Completing the Domestic Deployment of Full-Size DeepSeek

Thread Content

Recently, it was learned from the News Office of Sinopec that since the release of the domestically developed deep reasoning large model DeepSeek, Sinopec promptly established a dedicated team for model deployment and evaluation. On February 5, DeepSeek was deployed in a domestically produced computing environment, integrated into the Great Wall Large Model application system, and rolled out in batches within the company. This deployment not only provides strong technical support for enterprises’ intelligent transformation, but also offers valuable insights into the application of artificial intelligence technologies in the industrial sector, thereby contributing to technological innovation and efficiency improvements within this industry. Efficient deployment, thorough testing, and intelligent analysis to facilitate the creation of high-quality industry datasets. During the deployment process, Sinopec employed advanced inference acceleration technologies, which nearly doubled the inference computing efficiency of the large models and significantly enhanced the model’s capability to support various applications. To verify the suitability of DeepSeek-R1 for the petrochemical industry, Sinopec developed the \"Petrochemical Industry Large Model Test Set (Reasoning Edition)\") to conduct comprehensive and in-depth tests on the model. Test results show that DeepSeek-R1 has a strong ability to understand knowledge in the petrochemical industry, with a high overall accuracy in its responses. In addition, Sinopec also utilizes the DeepSeek large model to conduct intelligent analysis of hundreds of industry standards and technical specifications, in order to explore the creation of high-quality industry datasets. Tests have shown that the DeepSeek large model achieves high accuracy in recognizing professional content such as mathematical formulas, chemical structure diagrams, and specialized charts in the petrochemical industry, enabling it to effectively support the development of industry-specific datasets and the training of industry models. Next, the group’s artificial intelligence technology framework will be further improved to enhance the quality of training large models for the petrochemical industry, thereby achieving a deep integration of advanced reasoning capabilities with industry-specific knowledge. At the same time, the high-quality programming capabilities of the DeepSeek large model will be fully utilized to improve the development efficiency of specialized models for seismic data processing, reservoir development optimization, chemical product research and development, and customer service, thereby further driving the petrochemical industry toward intelligence and digital transformation. In recent years, Sinopec has made vigorous efforts to advance intelligent transformation and digitalization across the entire industrial chain, actively developing intelligent manufacturing, building artificial intelligence platforms, and deepening their application in specific scenarios. To date, 16 smart factories, 4 smart oil and gas fields, over 150 smart gas stations, and 3 intelligent research institutes have been established. 10 entities have been recognized as **-level pilot projects for intelligent manufacturing, while 7 companies have been awarded the **Excellence Level for smart factories, thereby driving continuous progress in the petroleum and petrochemical industry toward higher levels of sophistication, intelligence, and sustainability.