- Microsoft reveals seven new AI models
Microsoft’s AI chief has unveiled a series of the company’s new AI models aimed at encouraging developers to continue pushing the boundaries of technology.
Mustafa Suelyman, CEO of Microsoft AI, took the stage at Microsoft Build 2026 to announce the new releases, but also explained more about the company’s reasoning in its AI development.
He noted that Microsoft’s AI work will always seek to support human workers and users, not replace them, as it seeks to create what the company calls “humanistic superintelligence.”
“Humanistic superintelligence”
“We are truly living in the most remarkable times,” Suleyman said, “since I started working in AI, the computing we use to train frontier models has increased 1 trillion-fold, or 12 orders of magnitude in just 15 years. It is now clear that a steady, exponential increase in computing leads to predictable advances in AI capabilities, and in the coming years, we will see three orders of magnitude of computing applied to training frontier models.
“Intelligence is now a function of computing, long linear scaling has become the norm, scaling laws clearly hold, and it is an extraordinary time in our industry.”
Suleyman said his company is working toward the goal of a “humanistic superintelligence” with next-generation AI capabilities explicitly designed “to serve humans, not replace them.”
“The type of AI we create really matters,” he added, “we need AI that puts humanity first, that always prioritizes human well-being and human progress. This is the core philosophy and motivation behind our superintelligence efforts at Microsoft, and it shapes everything we do.”
Continuing his goal of “keeping developers building at the absolute frontier,” Suleyman announced no less than seven new Microsoft AI models in the fields of imaging, voice, and transcription.
But it also includes coding, with Microsoft’s new MAI-Code-1 model designed specifically for GitHub, and MAI-Thinking-1, its first reasoning model, trained using what Suleyman called “clean, commercially licensed data,” with high performance and low token cost.
To help spur the use and development of AI in scientific fields, the company also revealed more about Microsoft Discovery, its agent AI platform focused on “a new era of research and development.”
Now widely available, including an app in preview, the Microsoft Discovery app uses specialized human-deployed agents to mimic the scientific method on large amounts of knowledge to generate hypotheses and validate theories in a continuous loop.
Microsoft says the tool will be especially useful for customers in highly demanding and regulated industries such as food and energy, who may now have the ability to use agent discovery for their research and development, greatly accelerating the research process across industries.
Follow TechRadar on Google News and add us as a preferred source to receive news, reviews and opinions from our experts in your feeds.
