Microsoft's New AI Image Tool Beats Nano Banana on This Key Task

Microsoft unveiled two new text-to-image AI models, MAI-Image-2.5 and MAI-Image-Flash, at its Build developer conference on Tuesday. The models are part of a broader set of seven next-generation AI tools, including a reasoning model called MAI-Thinking-1 and updated voice, transcription, and coding models. MAI-Image-2.5 focuses on high-fidelity, professional-grade image generation, while the Flash variant prioritizes speed for production workloads. According to benchmarking on the Arena AI leaderboard, Microsoft’s MAI-Image-2.5 outperformed Google’s Nano Banana 2 in image editing tasks, though it still trails OpenAI’s GPT-Image-2 for overall performance. The new models are now integrated into PowerPoint and Microsoft’s enterprise marketplace, Foundry, with a broader rollout planned for OneDrive. Microsoft also highlighted the models’ precision in editing, offering users greater control and consistency compared to competitors. The company’s AI CEO, Mustafa Suleyman, emphasized their utility in creative and professional workflows during the Build keynote. The release marks Microsoft’s push into AI-driven creativity, expanding beyond its Copilot offerings. While Nano Banana 2 remains dominant in general image generation, Microsoft’s integration with its existing software ecosystem—such as PowerPoint and enterprise tools—could influence adoption for commercial and professional users. The new models are part of Microsoft’s broader AI strategy, which includes advancements in agentic AI and reasoning capabilities. Accessibility remains a key factor, with Microsoft’s tools tightly linked to its productivity suite, potentially giving it an edge for users already invested in the company’s ecosystem.