Microsoft tests AI models such as DeepSeek to compete for OpenAI

Microsoft is actively planning to enter the field of AI and strives to compete with OpenAI in the artificial intelligence track. It is reported that Microsoft is evaluating AI models developed by many companies such as Anthropic, DeepSeek and Meta, and simultaneously develops its own artificial intelligence inference model, which may be sold to developers in the future. In addition, Microsoft has tested AI technologies provided by companies such as Musk's xAI to enhance its Copilot products.

The rapid development of the DeepSeek model has attracted much attention. Among the 2025 ranking of the top 100 global generic AI applications released by Andreessen Horowitz (a16z), the DeepSeek web application ranks second, and its potential cannot be underestimated.

Recently, new news has also been reported by the highly-watched AI agent Manus. Its official "X" account has returned to normal after a brief freeze. Manus is positioned as an “autonomous agent”, able to complete complex tasks independently and deliver results directly, and surpass OpenAI similar products in GAIA benchmarks to reach SOTA levels.

What makes Manus unique is its "hands and brains" feature. Unlike ChatGPT, which only provides text generation, Manus can independently plan and execute tasks. For example, if users only need to upload a resume compression package, Manus can automatically decompress, analyze and generate recommendation reports without manual intervention and support asynchronous task processing.

The above is the detailed content of Microsoft tests AI models such as DeepSeek to compete for OpenAI. For more information, please follow other related articles on the PHP Chinese website!