Microsoft Reveals Seven New In-House AI Models at Build

TL;DR. Microsoft unveiled seven homegrown AI models at its Build conference, including a reasoning model matching Anthropic's Claude Sonnet 4.6. - The move signals Microsoft's push for greater AI self-sufficiency beyond key partners. - New models include MAI-Thinking-1, MAI-Code-1-Flash, and MAI-Image-2.5. - These cover reasoning, coding, and image generation, trained without external model distillation.

• Microsoft introduced seven new in-house AI models at its Build conference.
• MAI-Thinking-1, a reasoning model, reportedly matches Anthropic's Claude Sonnet 4.6 in testing and Claude Opus 4.6 on a coding benchmark.
• Microsoft also released MAI-Code-1-Flash for coding and MAI-Image-2.5 for image editing.
• This initiative aims to reduce Microsoft's reliance on external AI model providers like OpenAI and Anthropic.
• The models were trained from scratch to ensure clean data lineage, appealing to enterprise clients.

Back to QLANKR News