
Microsoft
Artificial Intelligence · Cybersecurity · Enterprise Software
Microsoft launches seven MAI models to undercut OpenAI, Anthropic on price
November 17, 2026
The strategy bets that task-specific models trained for narrow jobs can match frontier quality while costing a fraction to run around the clock.
- Microsoft unveiled seven new MAI models at its Ignite 2026 conference in San Francisco, covering coding, cybersecurity, image editing, real-time voice, transcription and decision-making.
- The models come from Microsoft AI (MAI), the in-house unit led by CEO Mustafa Suleyman, as the company pushes to build more of its own AI technology instead of relying solely on OpenAI and Anthropic.
- MAI Code 2 hits 98% of GPT-6.1 Sol's coding score at about 40% of the price, and 96% of Claude Opus 5.5's score at roughly 5% of the cost, per Microsoft's own benchmarks.
- MAI Cyber 2 Flash matches GPT-6 Astra's quality on the CyberGymE2E security benchmark at under 2% of the cost per task and is rolling out to a limited group of vetted defenders through Microsoft's MDASH system.
- Maia 200, Microsoft's second-generation AI accelerator chip, is now in production and delivers 1.7x more tokens per watt than its predecessor.
- Dragon Copilot Transcribe, built on the new MAI Transcribe 2 model, already processes 10 million patient encounters a month with multilingual error rates cut in half.
- Betting on cheap, specialized models rather than one giant general-purpose system reflects an industry-wide shift toward price-performance as AI agents run continuously and token costs climb.