Microsoft launches MAI-Thinking-1, a 35-billion-parameter reasoning AI model

Microsoft has released MAI-Thinking-1, a 35-billion-parameter AI model designed for reasoning tasks. It scores 97% on the AIME benchmark and 53% on SWE Bench Pro, outperforming Anthropic's Claude Opus 4.6 at 51.9% and surpassing OpenAI's GPT-5.4 on SWE Bench Pro at 59.1%. The model is described as Microsoft's first reasoning model and is positioned to compete with top-tier AI systems in complex coding and mathematical problem-solving. Early independent testers report it as superior to Claude Sonnet 4.6 in overall quality. This marks a rapid advancement for Microsoft, with its first in-house models launched less than a year ago. The release is part of a broader strategy to expand AI capabilities in enterprise environments through targeted model performance and competitive positioning.