Microsoft, 'Set Free' From OpenAI, Chases Superintelligence With Its Own Models
Microsoft is stepping out of OpenAI's shadow. The company's AI division has disclosed that a change to its contract with OpenAI, roughly six months ago, granted it formal authority to pursue 'superintelligence' using its own researchers, data pipelines and custom silicon.
That shift marks a significant evolution in one of the technology industry's most important and complicated partnerships. Microsoft has been OpenAI's largest backer and a primary channel for its models, but it has also been building considerable AI muscle of its own. Its AI chief has described the division as having been 'set free' to chart an independent path toward the most advanced systems.
The first fruits are already here. Microsoft's superintelligence team has released a family of seven in-house models, beginning with MAI-Thinking-1, described as the company's first reasoning model. It is a mid-sized system — 35 billion active parameters with a 256,000-token context window — built with an emphasis on efficiency as much as raw capability.
The move signals that Microsoft intends to compete at the frontier on its own terms rather than depending solely on a partner, hedging against the risk of relying on any single outside lab. It also intensifies an already crowded race in which OpenAI, Google, Meta and Anthropic are all pushing toward ever more capable models.
For Microsoft, owning the full stack — models, data and chips — offers strategic control over the technology it is weaving through its products. Whether 'superintelligence' proves an achievable goal or an aspirational label, the company is now clearly determined to build toward it under its own banner.
