Microsoft unveiled seven new AI models at its annual Build event on Tuesday, claiming they outperformed Anthropic's Claude Sonnet 4.6 and Google's Nano Banana 2 in blind testing and image-editing benchmarks. The launch represents Microsoft's effort to establish itself as a frontier AI developer rather than solely OpenAI's largest backer and infrastructure provider. The announcement comes as competition among leading AI developers continues to intensify, with Anthropic and Google recently releasing their own flagship models.
MAI-Thinking-1 Outperforms Claude Sonnet 4.6 in Blind Tests
MAI-Thinking-1, Microsoft's flagship text foundation model, was preferred over Anthropic's Claude Sonnet 4.6 in blind tests conducted by independent evaluators, according to Microsoft AI CEO Mustafa Suleyman. The model scored 97% on AIME 2025, a benchmark measuring advanced problem-solving and reasoning skills. Suleyman stated the model's SWE Bench Pro result places it "right alongside Opus 4.6 on one of the toughest coding benchmarks." Microsoft also claimed MAI "delivered the highest win rate, outperforming GPT-5.5 on quality, while being 10x lower on cost."
"Super excited to announce seven new world-class MAI models today," Suleyman wrote on X. "They represent what we consider a new era in AI designed to keep you in control and on the frontier."
Microsoft Launches Six Additional AI Models Across Coding, Image, Voice Categories
Microsoft introduced MAI-Code-1-Flash, a lightweight coding model built for GitHub Copilot and Visual Studio Code. The company also unveiled MAI-Image-2.5 and its Flash variant, which Microsoft says outperform Google's Nano Banana Pro on image-editing tasks. MAI Transcribe-1.5, a transcription model supporting 43 languages, and MAI-Voice-2, a speech-generation model capable of producing natural-sounding voices in 15 languages and adapting to a speaker from a short audio sample, were also announced.
"This is an extraordinary time in technology. The compute used to train frontier models has increased by a factor of one trillion," Suleyman said in a blog post announcing the new models. "Now we expect another thousand-fold increase over the next three years, which in turn means more advanced capabilities, and the continued rollout of ever more effective AI."
Anthropic and Google Release Competing Models
Last week, Anthropic announced the launch of Opus 4.8, which the company said is faster and smarter on benchmark tests and comes with new features. On Tuesday, Anthropic announced an expansion of its Project Glasswing, giving 150 companies access to its cybersecurity-focused Mythos model. At Google I/O in May, Google unveiled Gemini Omni, a multimodal AI model combining Gemini with Veo, Nano Banana, and Genie media-generation models, alongside Gemini Spark, a cloud-based AI agent designed to manage tasks across apps and workflows.
"Developers and businesses have been crying out for AI that delivers on their terms and under their say," Suleyman wrote. "We see this as a major step towards delivering that."
FAQ
What did Microsoft announce at Build event on Tuesday?
Microsoft unveiled seven new AI models including MAI-Thinking-1, MAI-Code-1-Flash, MAI-Image-2.5, MAI Transcribe-1.5, and MAI-Voice-2, claiming they outperformed Anthropic's Claude Sonnet 4.6 and Google's Nano Banana 2 in various benchmarks.
How did MAI-Thinking-1 perform on benchmarks?
MAI-Thinking-1 scored 97% on AIME 2025 and was preferred over Claude Sonnet 4.6 in blind tests conducted by independent evaluators, according to Microsoft AI CEO Mustafa Suleyman.
