Tencent Cloud TI Platform Unveils DeepSeek Series Models

Recently, Tencent Cloud's TI platform has officially launched the highly anticipated DeepSeek series models, which include the "full version" V3 featuring 671 billion parameters, the original R1 model, and several distilled models with parameter sizes ranging from 70 billion to 1.5 billion. This initiative aims to provide robust support for developers utilizing artificial intelligence (AI) tools, thereby promoting the widespread application of large model technology.

The DeepSeek series has garnered significant global attention due to its impressive performance metrics. Notably, the DeepSeek-R1 was released as an open-source model and extensively utilized reinforcement learning techniques during its post-training phase. This approach has dramatically enhanced the model's inference capabilities, even with limited labeled data. In various tasks, including mathematics, coding, and natural language reasoning, the performance of DeepSeek-R1 has been found to be comparable to that of OpenAI's GPT-4. Furthermore, DeepSeek-R1 operates under the MIT License, enabling users to train additional models through the distillation process. Its distilled counterpart, DeepSeek-R1-Distill, has also demonstrated remarkable performance in benchmark tests while possessing a smaller parameter size and reduced inference costs.

The Tencent Cloud TI platform not only facilitates one-click deployment of the DeepSeek series models but also offers a limited-time free online experience of the R1 model. This allows developers to engage with the models without barriers. Users can access the DeepSeek series model card in the "TI Platform - Large Model Plaza" to explore model details and participate in online experiences and one-click deployments. Additionally, the TI platform provides enterprise-level capabilities, including model service management, operational monitoring, and resource scaling, which assist businesses and developers in integrating DeepSeek models into real-world applications effectively and reliably.

To cater to the diverse needs of its user base, the TI platform has introduced various billing models, such as pay-as-you-go and annual/monthly subscriptions. Users seeking short-term experiences can opt for the pay-as-you-go model by purchasing computing power directly from the TI platform. For those who have already acquired Cloud Virtual Machines (CVM) or require long-term usage, it is advisable to utilize their existing CVM machines for inference tasks. In terms of optimal computing power configuration, deploying the "full version" DeepSeek-R1 is recommended on two 8-card HCCPNV6 models on Tencent Cloud for a stable operational experience. Conversely, the distilled DeepSeek-R1-Distill-Qwen-1.5B model can be effectively deployed on a single mid-range GPU card. Developers are encouraged to select appropriate models based on their business complexity and integrate them into AI applications via API calls.

This recent development by the Tencent Cloud TI platform not only equips developers with powerful AI tools but also significantly advances the popularization and application of large model technology. By offering free experiences and streamlined deployment features, the TI platform effectively lowers barriers for developers, enabling quicker application of AI technology to real-world business scenarios. This initiative further enhances the practicality and accessibility of AI technology for a broader audience.

Key Points

1. Tencent Cloud launches DeepSeek series models, including V3 and R1.
2. Models feature impressive performance and support various applications in AI.
3. The platform offers free trials and one-click deployment for developers.
4. Flexible billing options cater to diverse user needs.
5. Enhanced enterprise capabilities facilitate real-world integration of AI technology.