Microsoft develops new AI models to compete with OpenAI

Microsoft is reportedly developing in-house AI reasoning models that it believes can compete with industry leaders including its partner OpenAI. 

According to The Information, which cites a person involved with the programme, the company has been testing other AI models from Meta, xAI and DeepSeek as potential replacements to OpenAI in Copilot. 

The Information’s report, titled ‘Microsoft’s AI Guru Plots a Future Without OpenAI’ alleges that Microsoft is looking to reduce its dependence on Open AI, with which it has collaborated since 2019.

The company’s AI division, headed up by Mustafa Suleyman, has been developing and training a family of AI models referred to internally as MAI, it says. 

These perform comparatively to leading industry models from the likes of Anthropic and OpenAI on common benchmarks, says the source, who requested anonymity when discussing the internal developments. 

Complex reasoning models

The Information says Microsoft’s testing and development has centred around how MAI performs on a variety of tasks, including powering its Copilot AI assistants. 

It is also developing reasoning models – AI models that use logical processes and chain-of-thought techniques to simulate human thinking, make decisions and draw conclusions by using inference mechanisms and knowledge representation. 

Reasoning models break down complex problems into smaller and more manageable steps through intermediate reasoning abilities. 

Unlike general-purpose LLMs, which may generate direct answers, reasoning models are specifically trained to show their work and follow a more structured ‘thought’ process. 

This makes them well suited for applications and industries requiring complex problem solving such as medical diagnosis in healthcare, scientific research, automated decision making in manufacturing and financial risk assessment. 

By developing its own reasoning models Microsoft could directly compete with OpenAI, says The Information’s report. 

OpenAI’s o-series reasoning models are currently used by Microsoft in its Copilot products. They are also available in the Azure AI Foundry model catalogue and AI Toolkit for VS Code. 

According to The Information, Suleyman’s team is already trialling swapping the MAI models for OpenAI’s in Copilot.

Microsoft and OpenAI

Microsoft began its relationship with OpenAI in 2019. Since, the pair have collaborated across AI supercomputing and research to advance and commercialise AI technologies. 

Through the strategic partnership, Microsoft has invested heavily in OpenAI providing access to its Azure cloud platform to power OpenAI’s R&D while integrating its AI technology into its business and consumer products. 

OpenAI’s AI models are primarily hosted and run in Microsoft’s Azure cloud platform, while Microsoft incorporates the AI firm’s technology in products including Microsoft Office, GitHub Copilot and Bing Search. 

The companies also collaborate on the R&D of new AI technologies including large language models and AI supercomputing. 

Speaking in a Microsoft Blog in January 2023, when Microsoft announced the extension of the partnership into a third phase, CEO and Chairman Satya Nadella said: “We formed our partnership with OpenAI around a shared ambition to responsibly advance cutting-edge AI research and democratise AI as a new technology platform. 

“In this next phase of our partnership, developers and organisations across industries will have access to the best AI infrastructure, models and toolchain with Azure to build and run their applications.”

OpenAI CEO Sam Altman said of the partnership extension: “The past three years of our partnership have been great. Microsoft shares our values and we are excited to continue our independent research and work toward creating advanced AI that benefits everyone.”

Expanding AI adoption across Microsoft

By developing its MAI models, Microsoft could ultimately lessen its reliance on OpenAI. 

In December 2023, Reuters reported that Microsoft had been adding internal and third-party AI models to its 365 Copilot product. 

Citing sources familiar with the work, it said the company was looking to “diversify from the current underlying technology from OpenAI and reduce costs”. 

While the inclusion of OpenAI’s GPT-4 model was a major selling point for 365 Copilot when it was launched in March 2023, Reuters’ source said Microsoft was training its own smaller models, including Phi-4 to make Copilot faster and more efficient. 

Speaking at the Morgan Stanley TMT Conference on Tuesday 4 March 2025, Microsoft’s Chief Financial Officer Amy Hood said of the two companies’ relationship: “We’re both successful when each of us are successful. I do think everybody is planning for what happens for a decade or two decades. 

“And that's important for both of us to do. And what's great is that we're building a really flexible fleet that can be used for any type of workload on a global basis. And we look forward to continuing to be their primary partner and be able to supply them through that agreement and structure through 2030. It's a good thing for both of us.”

Explore the latest edition of AI Magazine and be part of the conversation at our global conference series, Tech & AI LIVE. 

Discover all our upcoming events and secure your tickets today.

AI Magazine is a BizClik brand