Product Launch read

Microsoft adds MAI-Code-2 and other in-house models at Ignite 2026

Microsoft announced new in-house models for coding, image generation and cybersecurity at Ignite, plus general availability of its Microsoft-Decision-1.1 routing model.

AE
Aidose Editorial
Nov 20, 2026

Microsoft used its Ignite 2026 conference, running November 17 to 20, to announce a set of new in-house AI models for coding, image generation and cybersecurity, according to the company's Ignite Book of News.

The coding models are MAI-Code-2, aimed at demanding agentic coding work, and MAI-Code-2-Flash, aimed at high-volume everyday coding where developers want to trade some performance for lower cost. Both are generally available in GitHub, and MAI-Code-2-Flash is in private preview in Microsoft Foundry.

Microsoft also announced MAI-Image-2.7 and MAI-Image-2.7-Flash, image generation models with improved editing and more consistent output, both in preview. MAI-Cyber-2-Flash, built for finding, validating and fixing code vulnerabilities, is in preview only within Microsoft's Codename MDASH.

Alongside the MAI models, Microsoft made Microsoft-Decision-1.1 generally available in Microsoft Foundry, OpenRouter and Vercel. The model handles routing, classification, prioritization and verification. Microsoft says it scores 68 on the Artificial Analysis Decision Index v2.1, six points higher than the first version released in October, runs more than twice as fast, and adds image inputs. It is built on Google's Gemma 4 26B A4B.

The releases reflect Microsoft's push to route enterprise workloads to smaller, purpose-built models instead of sending every request to a large general model. For teams on GitHub and Foundry, the practical change is more first-party model options for coding and request routing, with availability that varies by model and is still in preview for several of them.

AE
Aidose Editorial
Nov 20, 2026 · read
Back to News