The dark satanic rumour mill has manufactured a hell-on-earth yarn claiming that Vole is developing a series of large language models dubbed MAI that reportedly rival offerings from OpenAI and Anthropic. 

Bloomberg sources claim that Microsoft’s new AI models have been tested internally to see if they can power the company’s Copilot assistants, a job currently handled by OpenAI’s tech. The results suggest MAI is competitive with the best in the business, setting the stage for Redmond to reduce its reliance on ChatGPT. 

While the full details remain under wraps, MAI may be running on Vole’s Maia 100 AI chips, unveiled last year. This could allow the company to cut costs further and move away from Nvidia’s notoriously expensive GPUs.

Microsoft is developing a separate LLM series optimised for reasoning tasks, a sign that it’s gunning for broader AI capabilities beyond general-purpose chatbots. 

This shift comes amid an evolving relationship between Microsoft and OpenAI. Despite pouring over $13 billion into the AI startup, Microsoft has quietly diversified its AI portfolio. The two firms recently amended their partnership terms, allowing OpenAI to host workloads on rival cloud providers.

Meanwhile, Vole has been experimenting with LLMs from Anthropic, Meta, DeepSeek, and xAI to see if they could slot into Copilot. 

In addition to MAI, Microsoft has already succeeded with its Phi series of language models, which use synthetic data training methods. The latest Phi-4 models, launched in February, include a 3.8-billion-parameter model designed for reasoning and a multimodal variant that reportedly matches some of GPT-4’s capabilities. 

If Microsoft manages to integrate MAI into Copilot, it could signal a major strategic shift—one in which Redmond no longer depends on OpenAI but instead builds a diverse AI ecosystem powered by multiple competing models. It’s yet another sign that the AI landscape is shifting rapidly, and Microsoft is ensuring it has plenty of options in play.