Microsoft is set to announce new AI models at its Build conference on Tuesday, alongside updates to Windows 11 for developers.
The stage is set for Microsoft at a key turning point, as it shifts from relying on OpenAI for its AI clout to building its own in-house models. It is expected to announce its first reasoning model at Build, MAI-Thinking-1. The company has reportedly not used any distillation methods when training the model, which should make it a more unique venture.
Microsoft has reportedly shifted efforts and resources toward returning to the leading edge of AI model development, with AI chief Mustafa Suleyman saying the company must be at that level by 2027. This would be one of the first major tests of where Microsoft stands in terms of sophistication.
Alongside the reasoning model, Microsoft is expected to show MAI-Image-2.5 and MAI-Image-2.5-Flash, the company's two image generators, according to The Verge's Tom Warren. Image generation has been a key focus for Microsoft, with its model recently matching Google's Nano Banana on key benchmarks.
Microsoft may also bundle several of its AI model services into a single Copilot "super app", which could be unveiled at Build. One of the services in this bundle may be Microsoft Scout, the company's OpenClaw alternative for enterprises, which it is actively working on.
Even though Microsoft has expanded the number of services with Copilot functionality, including adding Copilot Work with Claude a few months ago, it has also been listening to feedback and reducing the amount of Copilot branding on Windows. It is expected to announce an even more scaled-back option for developers through an optimized experience with zero distractions.
Windows 11 will also receive performance improvements as part of Microsoft's ongoing effort to make the operating system more efficient.
New Microsoft hardware at the Build event
Microsoft will also showcase new hardware at Build, following Nvidia's announcement of the RTX Spark chip on Monday. During the keynote, Nvidia CEO Jensen Huang revealed the Surface Laptop Ultra, which is set to be Microsoft's ultimate laptop for AI workloads.
The RTX Spark chipset combines a microprocessor with a graphics chip and unified memory, and early benchmarks show it is a step up over the competition. It is the company's first major update for the consumer PC crowd in a while, as it has been focused on delivering new high-end chips for data centers and AI neoclouds.
Nvidia said Dell, Lenovo, HP, and ASUS will all launch devices with the chipset later in the year.
For Microsoft, RTX Spark has apparently been a three-year effort with Nvidia to build a chipset capable of delivering strong AI performance in a much smaller package. The announcement focused primarily on what these chips can deliver for AI workloads, with Nvidia claiming they can localize AI agent operations to ensure these agents do not operate outside core workloads.
Also read: Microsoft and Nvidia’s RTX Spark push brings Windows on Arm, Surface hardware, and local AI agents into the next AI PC race.