Tencent Cloud TI platform officially launched the DeepSeek series of large models, including the "full blood version" V3 and R1 original models with a parameter volume of 671B, as well as multiple distillation models. With its excellent performance and MIT open source license, DeepSeek-R1 performs outstandingly in tasks such as mathematics, code, natural language reasoning, and even comparable to GPT-4. This time, it not only provided developers with strong AI tools, but also reduced the use threshold of large models, and promoted the popularization and application of large model technology.
Recently, Tencent Cloud TI platform announced that it had officially launched the highly anticipated DeepSeek series models, including the "full blood version" V3 and R1 original models with a parameter volume of 671B, as well as the series models based on DeepSeek-R1 distillation, with parameter scales ranging from 70B to 1.5B ranges. This measure provides developers with strong AI tool support, which further promotes the popularization and application of large model technology.
The DeepSeek series model has attracted widespread attention globally with its excellent performance. Among them, DeepSeek-R1 was open sourced when it was released, and reinforcement learning technology was used on a large scale in the post-training stage. Even if there is only very little labeled data, it can greatly improve the model inference ability. In mathematics, code, natural language reasoning and other tasks, the performance of DeepSeek-R1 is comparable to the official version of the OPENAI GPT-4. In addition, DeepSeek-R1 follows the MIT License, allowing users to train other models through distillation technology. Its distillation model DeepSeek-R1-Distill still performs well in benchmarks with smaller parameter scales and lower inference costs.
Tencent Cloud TI platform not only fully supports the one-click deployment of the DeepSeek series models, but also opens the free online experience of the R1 model for a limited time, providing developers with zero threshold unboxing experience. Users can click the DeepSeek series model card in the "TI Platform-Big Model Plaza" to understand the model information and perform an online experience and one-click deployment. In addition, the TI platform also provides enterprise-level capabilities such as model service management, monitoring and operation, and resource scaling to help enterprises and developers efficiently and stably connect the DeepSeek model to actual business.
In order to meet the needs of different users, the Ti platform provides a variety of billing models, including billing and packets. For users who need a short-term experience, they can directly purchase computing power from the TI platform and choose the meter-by-quantity billing mode; for users who have purchased CVM machines or need a long-term experience, it is recommended to use their own CVM machines as their inference computing power Essence In terms of computing power configuration, the "full blood version" DeepSeek-R1 is recommended to use two 8-card HCCPNV6 models on Tencent Cloud for service deployment to obtain a stable business experience; and the distilled DeepSeek-R1-Distill-Qwen-1.5 The B model can be deployed in a single mid -end GPU card. Developers can choose the appropriate model for testing according to business complexity, and connect them to AI applications by calling API.
This new move by Tencent Cloud TI platform not only provides developers with powerful AI tool support, but also provides a strong boost to the popularization and application of large-scale model technologies. Through free experience and one-click deployment functions, the TI platform lowers the threshold for developers to use big models, allowing them to apply AI technology to actual business more quickly, further improving the practicality and accessibility of AI technology.
In short, Tencent Cloud TI platform launched the DeepSeek series models, providing developers with convenient and efficient large-scale application experience, further promoting the development and application of artificial intelligence technology, and is worth looking forward to its future development.