In an industry where AI leaders rarely concede ground, Mustafa Suleyman, CEO of Microsoft AI, has done just that.
Speaking to Bloomberg, Suleyman admitted that Google’s latest model has capabilities Copilot does not.
“Gemini can do things that Copilot can’t do,”
he said, before adding that Copilot also offers features Gemini lacks. While carefully balanced, the statement stands out in a sector dominated by claims of superiority rather than nuance.
Table of Contents
What Gemini does better
The model in question is Gemini 3, which Google has positioned as its most advanced multimodal system so far.
Gemini’s strength lies in raw capability. It focuses on understanding complex prompts, combining multiple data types, and generating creative or analytical outputs at a high level. According to Suleyman, this emphasis allows Gemini to outperform Copilot in specific technical and reasoning tasks.
That approach aligns with Google’s broader goal of building the most capable general purpose AI model rather than a tightly integrated assistant.
Why Microsoft thinks Copilot still wins
Suleyman was quick to reframe the comparison by stressing what Microsoft Copilot is designed to do best.
“Copilot is actually amazing for vision,” he said. “It can see everything that you are seeing and talk to you in real time.”
Copilot’s standout feature is its ability to understand live visual context. Users can share their screen on mobile or desktop and receive immediate feedback, making it more practical for day to day problem solving.
Rather than chasing benchmark dominance, Microsoft is prioritizing usefulness inside real workflows.
Copilot everywhere, not just in a lab
That philosophy explains why Copilot is being pushed deeply into Microsoft’s ecosystem.
It is now embedded across Windows 11, Outlook, Excel, and Microsoft Edge, where a dedicated Copilot mode offers in browser assistance. The idea is constant availability, not occasional brilliance.
Suleyman described the goal as building an assistant that “unblocks you whenever you get stuck,” whether that means editing a document, understanding a spreadsheet, or navigating a complex interface.
Humanist AI, not runaway intelligence
Suleyman also reiterated Microsoft’s broader stance on AI development, which the company calls “humanist superintelligence.”
The concept centers on keeping humans firmly in control. Suleyman said Microsoft would abandon any system that showed signs of unpredictable or autonomous behavior.
“We won’t continue to develop a system that has the potential to run away from us”
he said, drawing a clear boundary between advanced assistance and independent agency.
Different tools for different users
Google and Microsoft are no longer chasing the same finish line.
Gemini aims to be the smartest model in the room, capable of complex reasoning and creative synthesis. Copilot is being shaped as a reliable companion that fits naturally into daily work.
Suleyman’s comments suggest a future where multiple AI models coexist, each excelling in different areas depending on what users actually need.
For once, the AI race looks less like a winner takes all contest and more like a growing toolbox.
