Microsoft ($MSFT) Unveils In-House MAI AI Model Suite at Ignite 2026, Challenging OpenAI on Price-Performance
Microsoft ($MSFT) unveiled a new suite of in-house Microsoft AI (MAI) foundation models spanning software development, cybersecurity, image editing, voice, and clinical transcription at its Ignite 2026 conference in San Francisco on Tuesday, November 17—the same day strategic partner OpenAI released GPT-6.2 Astra and GPT-6.2 Sol.
Led by Microsoft AI chief Mustafa Suleyman, the rollout underscores Redmond's push to build domain-specialized, cost-efficient proprietary models alongside its ongoing work on a general-purpose frontier system:
- Coding and software engineering: Generally available in GitHub Copilot across VS Code, Visual Studio, and Copilot CLI, MAI Code 2 delivers 98% of GPT-6.1 Sol's coding performance at roughly 40% of the cost, while lightweight variant MAI Code 2 Flash outperformed GPT-6.1 Sol in AT&T enterprise evaluations at one-fiftieth (2%) of the expense.
- Cybersecurity and safeguard focus: Offered initially to a vetted group of enterprise security customers, MAI Cyber 2 Flash outperformed OpenAI's GPT-6 Astra on cybersecurity benchmarks at less than 2% of the per-task cost, directly targeting AI-driven software vulnerabilities and automated threat triage.
- Multimodal voice, image, and medical applications: MAI Realtime 1 topped GPT Realtime 2.1 across public and internal voice benchmarks, MAI Image 2.7 entered public preview at the top of Artificial Analysis image-editing rankings, and MAI Transcribe 2 debuted inside Dragon Copilot to power specialized medical and clinical documentation workflows.