Microsoft says its latest AI models beat Google’s Cloud and Nano Banana

On the first day of its annual Microsoft Build event on Tuesday, the Windows developer unveiled seven new AI prototypes, claiming they outperformed Anthropic’s Claude Sonnet 4.6 and Google’s Nano Banana 2 in blind tests and photo editing benchmarks.

This claim comes as Microsoft tries to establish itself as a leading AI developer rather than just the largest supporter and infrastructure provider for OpenAI.

“I’m very excited to announce seven new world-class MAI models today,” said Microsoft AI CEO Mustafa Suleiman books On X. “It represents what we see as a new era in AI designed to keep you in control and at the limits.”

At the center of the release is MAI-Thinking-1, a logical model that Microsoft describes as the main body model.

According to Solomon, the MAI-Thinking-1 was preferred over Anthropic’s Claude Sonnet 4.6 in blind tests conducted by independent raters. He added that the model received a score of 97% in AIME 2025, a standard that measures advanced problem-solving and thinking skills.

The SWE Bench Pro score puts the model “alongside Opus 4.6 on one of the toughest programming benchmarks,” Solomon said.

The company also introduced MAI-Code-1-Flash, a lightweight coding model designed for GitHub Copilot and Visual Studio Code; MAI-Image-2.5 and its Flash variant, which Microsoft says outperforms Google’s Nano Banana Pro at image editing tasks; MAI Transcribe-1.5, a transcription model that supports 43 languages; and MAI-Voice-2, a speech generation model capable of producing natural sounds in 15 languages ​​and adapting to the speaker from a short audio sample.

“This is an extraordinary time in technology. The computation used to train frontier models has increased by a factor of a trillion,” Soliman said in a separate blog. mail Announcing new models. “Now we expect another thousand-fold increase over the next three years, which in turn means more advanced capabilities and the continued deployment of ever more effective AI.”

This announcement comes as competition continues to intensify among top artificial intelligence developers.

Last week, Anthropic announced the launch of its latest flagship model, Opus 4.8which the company said is faster and smarter in benchmark tests and comes with a host of new features. Anthropic announced Tuesday expansion From its project Glasswing, giving 150 companies access to the new Mythos model focused on cybersecurity.

Meanwhile, at Google I/O in May, Google unveiled it Gemini Omnia multi-media AI model that combines Gemini with the company’s Veo, Nano Banana and Genie media generation models, along with… Gemini Sparka cloud-based AI agent designed to manage tasks across applications and workflows on behalf of the user.

Microsoft’s launch of the new model signals a broader effort to build proprietary AI systems as it expands beyond its long reliance on OpenAI technology, saying that MAI “delivered the highest win rate, outperforming GPT-5.5 in terms of quality, with a 10x lower cost.”

“Developers and companies have been crying out for AI that lives up to their terms and under their word,” Soliman wrote. “We see this as a big step towards achieving that.”