Microsoft’s In-House AI Models Challenge OpenAI
Microsoft Corp. is developing in-house artificial intelligence (AI) models that it believes can rival industry leaders, including its partner OpenAI, according to a source familiar with the matter. This move signifies a growing ambition within Microsoft to compete at the forefront of AI innovation.
A family of models, internally dubbed MAI, recently yielded impressive test results. These results suggest the models can compete with cutting-edge AI rivals, including products from OpenAI and Anthropic. The source, who requested anonymity to discuss internal matters, noted the competitive performance.
Microsoft has been rigorously testing the MAI models across various tasks, including powering components of the company’s Copilot-branded AI assistants. Copilots are designed to assist users with a diverse range of queries, while also offering tailored suggestions, especially for tasks like document creation or managing conference calls.
Furthermore, Microsoft is also advancing so-called reasoning models, which are engineered to address intricate queries and exhibit human-like problem-solving capabilities. OpenAI and Anthropic, along with Alphabet Inc., are also working on such models, illustrating a broader industry trend.
Notably, Microsoft integrated OpenAI’s o1 reasoning model into its Copilot products last month, showcasing a blended approach. “As we’ve said previously, we are using a mix of models, which includes continuing our deep partnership with OpenAI, along with models from Microsoft AI and open source models,” stated a Microsoft spokesperson, reflecting the company’s strategy.
Microsoft’s internal development efforts, first reported by The Information, could decrease their reliance on OpenAI. Microsoft has provided substantial backing to the ChatGPT creator, investing approximately US$13 billion (RM57.40 billion). Considering the size of the investment, the relationship between the two companies is subject to ongoing industry discussion.
The two companies recently adjusted their deal, announcing in January that OpenAI could use servers from other cloud-computing rivals to power its services, provided Microsoft didn’t want the business itself. Their agreement lasts until 2030.
OpenAI declined to provide comment on the matter. Microsoft’s chief financial officer, Amy Hood, emphasized the collaborative nature, stating at a Morgan Stanley conference earlier this week, “We’re both successful when each of us are successful. So as you go through that process, I do think everybody’s planning for what happens for a decade, or two decades. And that’s important for both of us to do.”
Microsoft currently offers a suite of smaller, internally built models, known as Phi, in addition to re-selling AI models developed by a wide range of other companies. The company evaluated various models from companies including Anthropic, DeepSeek, Meta Platforms Inc., and Elon Musk’s xAI to support Copilot, according to The Information.
“We feel great about having leading models from OpenAI, we’re still incredibly proud of that,” Hood previously said. “But we also have other models, including ones we build, to make sure that there’s choice.”
