Recently, Tencent Cloud's TI platform announced the official launch of the highly anticipated DeepSeek series models, including the "full version" V3 with 671B parameters and the original R1 model, as well as a series of models distilled from DeepSeek-R1, with parameter sizes ranging from 70B to 1.5B. This initiative provides developers with powerful AI tool support and further promotes the popularization and application of large model technology.
The DeepSeek series models have gained widespread attention globally due to their outstanding performance. Notably, DeepSeek-R1 was open-sourced upon release and utilized reinforcement learning techniques extensively during the post-training phase, significantly enhancing model inference capabilities even with minimal labeled data. In tasks such as mathematics, coding, and natural language reasoning, the performance of DeepSeek-R1 is comparable to that of OpenAI's official GPT-4. Additionally, DeepSeek-R1 follows the MIT License, allowing users to train other models through distillation. Its distilled model, DeepSeek-R1-Distill, demonstrates excellent performance in benchmark tests while having a smaller parameter size and lower inference cost.
The Tencent Cloud TI platform not only fully supports one-click deployment of the DeepSeek series models but also offers a limited-time free online experience of the R1 model, providing developers with a no-barrier experience. Users can click on the DeepSeek series model card in the "TI Platform - Large Model Plaza" to learn about the model and engage in online experiences and one-click deployments. Moreover, the TI platform provides enterprise-level capabilities such as model service management, operational monitoring, and resource scaling, assisting businesses and developers in efficiently and stably integrating DeepSeek models into real-world applications.
To meet the diverse needs of users, the TI platform offers various billing models, including pay-as-you-go and annual/monthly subscriptions. Users who need short-term experiences can directly purchase computing power from the TI platform and choose the pay-as-you-go model; for those who have already purchased CVM machines or require long-term experiences, it is recommended to use their own CVM machines for inference power. In terms of computing power configuration, the "full version" DeepSeek-R1 is recommended to be deployed on two 8-card HCCPNV6 models on Tencent Cloud for a stable business experience, while the distilled DeepSeek-R1-Distill-Qwen-1.5B model can be deployed on a single mid-range GPU card. Developers can select appropriate models for testing based on the complexity of their business and integrate them into AI applications via API calls.
This new move by the Tencent Cloud TI platform not only provides developers with powerful AI tool support but also significantly promotes the popularization and application of large model technology. Through free experiences and one-click deployment features, the TI platform lowers the barriers for developers to use large models, enabling them to apply AI technology to real-world business scenarios more quickly, further enhancing the practicality and accessibility of AI technology.
