Microsoft has created its own large language models that, according to the software giant, can compete with similar models from industry leaders, including OpenAI, Bloomberg writes, citing its own informed sources.
Image source: BoliviaInteligente / Unsplash
Microsoft’s AI models have recently been tested and found to be competitive with state-of-the-art models, including those from OpenAI and Anthropic. Microsoft’s AI algorithms, known collectively as MAI, are reportedly capable of performing a variety of tasks, including powering parts of its Copilot-branded AI assistants.
The source said that Microsoft is developing a “reasoning” AI model that can handle complex queries and perform various tasks. Other companies, such as OpenAI, Anthropic, and Google, are also developing “reasoning” AI models. Last month, Microsoft began using OpenAI’s o1 “reasoning” model in its Copilot products.
Microsoft’s own AI models could reduce the company’s dependence on OpenAI algorithms in the future. Recall that Microsoft invested about $13 billion in the ChatGPT developer. The companies recently revised their agreements and announced at the beginning of the year that OpenAI would be able to use third-party cloud platforms for its services if Microsoft did not want to engage in this business itself. The current agreement between OpenAI and Microsoft is valid until 2030.
Microsoft already offers its customers a suite of its small Phi AI models, as well as algorithms from other companies. According to the source, the company has tested how algorithms from various developers, including Anthropic, DeepSeek, Meta✴ Platforms, and xAI, would work as the basis for Copilot.