Microsoft has announced seven new artificial intelligence models, asserting that its flagship reasoning model outperformed Anthropic's Claude Sonnet 4.6 in blind evaluations and matched Claude Opus 4.6 on a leading coding benchmark. The company also said its image models surpassed Google's Nano Banana 2 on image-editing leaderboards.
The launch, made on the first day of the annual Microsoft Build conference on Tuesday, marks the company's most ambitious push yet to develop proprietary frontier AI models alongside its longstanding partnership with OpenAI.
"Super excited to announce seven new world-class MAI models today," Microsoft AI CEO Mustafa Suleyman wrote on X. "They represent what we consider a new era in AI designed to keep you in control and on the frontier."
At the center of the release is MAI-Thinking-1, a reasoning model that Microsoft describes as its flagship text foundation model. According to Suleyman, the model was preferred over Anthropic's Claude Sonnet 4.6 in blind tests conducted by independent evaluators. He added that MAI-Thinking-1 scored 97% on AIME 2025, a benchmark that measures advanced problem-solving and reasoning skills. Suleyman also said the model's SWE Bench Pro result places it "right alongside Opus 4.6 on one of the toughest coding benchmarks."
The full lineup of new models includes:
MAI-Thinking-1 — the flagship reasoning and text foundation model
MAI-Code-1-Flash — a lightweight coding model built for GitHub Copilot and Visual Studio Code
MAI-Image-2.5 and its Flash variant — image-editing models that Microsoft claims outperform Google's Nano Banana Pro on image-editing tasks
MAI Transcribe-1.5 — a transcription model supporting 43 languages
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
"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."
Microsoft also highlighted competitive cost advantages, stating that MAI "delivered the highest win rate, outperforming GPT-5.5 on quality, while being 10x lower on cost."
"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."
The announcement comes amid intensifying competition among leading AI developers. Last week, Anthropic announced the launch of its latest flagship model, Opus 4.8, which the company said is faster and smarter on benchmark tests and comes with a suite of new features. On Tuesday, Anthropic also announced an expansion of its Project Glasswing, giving 150 companies access to its new cybersecurity-focused Mythos model.
Meanwhile, at Google I/O in May, Google unveiled Gemini Omni, a multimodal AI model that combines Gemini with the company's Veo, Nano Banana, and Genie media-generation models, alongside Gemini Spark, a cloud-based AI agent designed to manage tasks across apps and workflows on a user's behalf.
Microsoft's new model launch signals a broader effort to build proprietary AI systems as it looks to expand beyond its longstanding reliance on OpenAI technology.
Why it matters
Microsoft's blind-evaluation methodology — using independent evaluators rather than self-reported scores — sets a higher bar for benchmark credibility claims in the AI industry.
The simultaneous release of seven models spanning reasoning, coding, image editing, transcription, and voice generation signals that Microsoft is competing across multiple AI capability categories at once, not just in a single domain.
The competitive landscape is shifting rapidly: within the same week, Anthropic launched a new flagship model and expanded a cybersecurity-focused program, while Google had already unveiled a multimodal model suite at its own developer conference.
According to the company, the launch reflects a strategic effort to build proprietary AI systems and reduce reliance on a single external technology partner.
