On the evening of March 25th, Tencent Cloud announced the launch of the DeepSeek-V3-0324 model API, allowing businesses and developers to access this latest model and receive stable, high-quality service directly through Tencent Cloud. Simultaneously, Tencent Cloud's large model knowledge engine has integrated the new DeepSeek-V3 model. Using the platform's built-in RAG (Retrieval-Augmented Generation) and workflow capabilities, users can quickly build their own AI applications.
Tencent Cloud offers two usage methods: directly calling the API on the Tencent Cloud website, or quickly building applications based on the new V3 model integrated into the large model knowledge engine. The pricing for the DeepSeek-V3-0324 API is comparable to the official DeepSeek pricing: 2 RMB per million input tokens and 8 RMB per million output tokens.
The new DeepSeek-V3 model shows significant improvements in several areas. It leverages reinforcement learning techniques used in the training of the DeepSeek-R1 model, significantly enhancing its performance on reasoning tasks. It achieved scores exceeding GPT-4.5 on mathematical and code-related benchmark datasets. In programming, the new V3 model boasts enhanced front-end development capabilities. The generated HTML code is not only more usable but also visually appealing and well-designed. Many developers believe its programming capabilities are comparable to Claude 3.5/3.7 Sonnet. In Chinese writing tasks, the new V3 model further optimizes writing quality, particularly improving the content quality of long-form texts. Furthermore, in online search scenarios, the new V3 model produces more detailed, accurate, and aesthetically pleasing results.
Tencent Cloud API:https://cloud.tencent.com/document/product/1772/115963
