Microsoft Corporation has stepped into the competitive landscape of large language models (LLMs) with a new AI series reportedly on par with industry leaders like OpenAI and Anthropic PBC.
According to multiple reports, the newly developed LLMs, known as MAI, signify Microsoft’s ongoing push for greater AI independence.
Sources familiar with the matter told Bloomberg that MAI stands for “Microsoft Artificial Intelligence,” though it may also reference Maia 100, the proprietary AI chip Microsoft introduced last year. It is speculated that the Maia 100 processor could be powering this new series of models.
Microsoft recently conducted internal tests on MAI to assess its capabilities. These evaluations aimed to determine whether the LLMs could effectively support Microsoft’s Copilot, the AI-powered assistant integrated across its suite of products.
Reports indicate that MAI performed competitively against leading models from OpenAI and Anthropic, suggesting that Microsoft’s investment in homegrown AI is paying off.
The fact that Microsoft tested MAI within Copilot indicates that the new model series is designed for general-purpose AI processing rather than highly specialized reasoning tasks. Copilot relies on LLMs to generate text, summarize information, and perform coding assistance—functions that general-purpose models excel at.
Microsoft has also reportedly been working on a separate LLM series optimized for reasoning tasks, though no details have been confirmed regarding its architecture or capabilities.
A Move Toward Greater AI Autonomy
If Microsoft integrates MAI into Copilot, it could mark a shift away from its reliance on OpenAI’s technology.
Microsoft has poured over $13 billion into OpenAI and, until recently, was its exclusive cloud provider. However, in January 2025, the terms of their agreement changed, allowing OpenAI to explore other cloud platforms.
This development suggests that Microsoft may be exploring alternative AI solutions for its products. In addition to MAI, the company has reportedly tested models from Anthropic, Meta, DeepSeek, and xAI Corp. as potential replacements or complementary options for Copilot.
“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,” a Microsoft spokesperson told Bloomberg.
Microsoft’s Growing LLM Portfolio
MAI is not Microsoft’s first venture into developing proprietary LLMs. The company has already released Phi, a series of open-source language models engineered for power efficiency. The Phi-4 series, launched in February, includes:
- Phi-4-mini: A 3.8 billion parameter model designed for reasoning-heavy tasks like solving complex math problems.
- Phi-4-multimodal: A more advanced version that can handle multimodal input, improving its ability to process images, text, and other data types simultaneously.
Microsoft has stated that Phi-4-multimodal achieves performance levels close to GPT-4, despite having significantly fewer parameters.
Microsoft’s development of MAI and the ongoing work on reasoning-optimized LLMs suggest the company is strategically positioning itself as a leader in AI innovation. The company’s synthetic data-driven LLM training methods, which played a crucial role in the development of Phi-4, could also be leveraged to enhance MAI’s capabilities.
As competition in the AI sector intensifies, Microsoft’s move toward developing proprietary AI models not only reduces its reliance on external providers like OpenAI but also positions it to offer more customized, cost-effective AI solutions for enterprise and consumer applications.
The full impact of MAI remains to be seen, but its introduction signals that Microsoft is fully committed to securing a dominant role in the future of AI-driven technology.
