Microsoft launches 7 MAI models and cuts pricing to rival Anthropic
Seven in-house MAI models aim to match Opus 4.6 at lower costs.
At Microsoft Build 2026, AI CEO Mustafa Suleyman unveiled seven new in-house MAI models, signaling the company's aggressive push to compete with Anthropic and other AI leaders. The lineup spans specialized capabilities: MAI-Image-2.5 for image generation, MAI-Transcribe-1.5 for speech-to-text, MAI-Voice-2 for text-to-speech, MAI-Thinking-1 for reasoning tasks, and MAI-Code-1-Flash for code generation. Additionally, the Aion 1.0 series introduces instruction-following (Instruct) and planning-focused (Plan) models.
Suleyman emphasized reduced pricing across all models, directly targeting Anthropic's Opus 4.6. Microsoft claims its new models deliver comparable performance for common enterprise use cases at significantly lower costs. This move is designed to attract price-sensitive developers and enterprises, especially as AI inference expenses remain a major barrier. By offering a broad spectrum of specialized models rather than a single giant, Microsoft aims to let customers choose the right tool for each task, potentially undercutting Anthropic's pricing while maintaining quality.
- Seven new MAI models launched: Image, Transcribe, Voice, Thinking, Code, Aion Instruct, and Aion Plan.
- Pricing reduced to undercut Anthropic's Opus 4.6, with comparable capabilities.
- Models target specific tasks: reasoning, code, speech, and planning for enterprise flexibility.
Why It Matters
Microsoft undercuts AI pricing with specialized models, forcing competitors to rethink cost structures.