Microsoft, Nvidia, Alibaba, Baidu are all competing for it. DeepSeek has become a mainstream standard.

In just one week, overseas technology giants such as Microsoft Azure and NVIDIA, as well as domestic cloud vendors such as Alibaba Cloud, Huawei Cloud, Tencent Cloud, and Baidu Cloud have announced the launch of the DeepSeek large model, and attracted more users with discounts such as “zero code” and “ultra-low price”.

Since 2025, DeepSeek has successively open-sourced the large language model V3, the inference model R1, and the multimodal model Janus Pro, which has taken the lead in attracting attention from overseas markets due to its advantages of high performance and low cost. As one of OpenAI’s “financiers”, Microsoft CEO Nadella praised DeepSeek’s innovation on January 29, and also pulled a wave of advertising for Microsoft itself: the DeepSeek-R1 model is available through Microsoft’s AI platform Azure AI Foundry and GitHub, and promises to run on computers with Copilot+ in the future.

On January 30th, NVIDIA also announced that its developer website has included the DeepSeek-R1 model in the “Most Popular Models” section, which can be used on the NVIDIA NIM microservice preview version. NVIDIA NIM is part of NVIDIA AI Enterprise Edition services, providing containers for self-hosted GPU-accelerated inference microservices across clouds, data centers, and workstations, for pre-training and custom AI models.

Soon, major domestic internet cloud service providers began to compete to access the DeepSeek series models and put forward ‘real money’ to compete with each other.

On February 3rd, Aliyun announced that Aliyun PAI Model Gallery supports one-click deployment of DeepSeek-V3 and R1 models on the cloud. On this platform, users can achieve the entire process from training to deployment to inference without writing a single line of code, simplifying the model development process, and bringing faster, more efficient, and more convenient AI development and application experience to developers and enterprise users.

On the same day, Baidu Intelligent Cloud Qianfan Platform also officially launched the DeepSeek-R1 and V3 models, introducing ultra-low price plans and offering limited-time free services. You can quickly experience it by logging into the Baidu Intelligent Cloud Qianfan ModelBuilder. At the same time, Baidu has integrated its own large-scale model technology, such as integrating the DeepSeek model into the Qianfan reasoning chain, integrating Baidu’s exclusive content security operator, and achieving enhanced security protection for models.

I learned that many domestic AI companies, including Huawei Cloud, Tencent Cloud, 360 Digital Security, and Cloud Axis Technology ZStack, have integrated the DeepSeek model, which is almost a standard configuration for mainstream AI cloud platform services.

While DeepSeek has been on a wild ride, the AI big model industry is gradually entering a rally.

On January 29th, Ali Tongyi launched the flagship model Qwen2.5-Max on Alitranslate. It adopts the ultra-large-scale MoE (Mixed Expert) architecture and is based on pre-training data of over 200 trillion tokens. The Ali team stated that the model has demonstrated comparable or even leading performance compared to DeepSeek V3, GPT-4, and Claude-3.5-Sonnet. With the continuous improvement of post-training technology, the next version is expected to reach an even higher level.

On February 1st, OpenAI launched o3-mini, which is OpenAI’s first small-scale inference model that supports developer-specific functionality. It has the advantages of low cost, low latency, and supports function calls, streaming transmission, structured output, and other functions. It also supports integration with search functions, providing the latest answers and linking to relevant network resources. After testing by some developers, it was found that the performance of OpenAI o3-mini is not significantly better than that of DeepSeek R1, and some aspects are still surpassed by R1.

(Article source: The Paper)

Source: Eastmoney.com

Author: Shangguan News