Microsoft introduced MAI-Thinking-1, a 35 billion active parameter reasoning model, at its Build 2026 developer conference. The model outperforms Anthropic's Claude Sonnet 4.6 in independent evaluations, according to the company.
TL;DR
- Microsoft launched MAI-Thinking-1, a 35B-parameter reasoning model, at Build 2026.
- The model outperforms Anthropic's Claude Sonnet 4.6 and matches Claude Opus 4.6 on the SWE Bench Pro coding benchmark.
- Six additional MAI models were announced, covering image generation, transcription, voice, and code.
What happened
Microsoft announced a new family of AI models, MAI, at its Build 2026 developer conference. The centerpiece is MAI-Thinking-1, a 35 billion active parameter reasoning model with a 128K context window. The model is designed for complex multi-step instructions, long context reasoning, and code generation. According to Kyle Daigle, Microsoft Developer CMO and COO of GitHub, independent evaluators preferred MAI-Thinking-1 over Anthropic's Claude Sonnet 4.6, and it matches Claude Opus 4.6 on the SWE Bench Pro coding benchmark.
In addition to MAI-Thinking-1, Microsoft announced six other MAI models: MAI-Image-2.5 and a Flash variant for image generation, MAI-Transcribe-1.5 for transcription, MAI-Voice-2 and a Flash variant for voice, and MAI-Code-1 for code. MAI-Thinking-1 is available in Microsoft Foundry as a private preview, while MAI-Image-2.5 models are already live in PowerPoint and OneDrive, and MAI-Code 1 is available now in Copilot and VS Code.
Why it matters
The launch of MAI-Thinking-1 and the other MAI models represents a significant step forward for Microsoft in the AI space. The models are designed to handle a wide range of tasks, from reasoning and coding to image generation and voice. This could make them valuable tools for developers and businesses looking to integrate AI into their workflows.
The performance of MAI-Thinking-1, in particular, is notable. By outperforming Anthropic's Claude Sonnet 4.6 and matching Claude Opus 4.6 on the SWE Bench Pro coding benchmark, Microsoft is positioning itself as a strong competitor in the AI model market. This could have implications for the broader AI landscape, as well as for developers and startups looking to leverage AI in their products and services.
Key facts
- MAI-Thinking-1 is a 35 billion active parameter reasoning model with a 128K context window.
- Independent evaluators preferred MAI-Thinking-1 over Anthropic's Claude Sonnet 4.6.
- MAI-Thinking-1 matches Claude Opus 4.6 on the SWE Bench Pro coding benchmark.
- Six additional MAI models were announced, covering image generation, transcription, voice, and code.
- MAI-Thinking-1 is available in Microsoft Foundry as a private preview.
- MAI-Image-2.5 models are already live in PowerPoint and OneDrive, and will be arriving soon in Foundry.
- MAI-Code 1 is available now in Copilot and VS Code.
- MAI-Transcribe-1.5 will be available soon in 43 languages, while MAI-Voice-2 and a Flash variant are already available in 15 additional languages with multiple voice options.
Context
The launch of the MAI models is part of a broader trend in the AI industry towards developing more powerful and specialized models. As AI becomes increasingly integrated into various aspects of business and daily life, the demand for models that can handle specific tasks is growing.
Microsoft's entry into the AI model market with the MAI family of models is a significant development. The company is already a major player in the tech industry, and its foray into AI models could have implications for the broader AI landscape. It will be interesting to see how the MAI models perform in the market and how they compare to other AI models from companies like Anthropic and OpenAI.