# Microsoft Announces Two MAI Code 2 Models for GitHub Copilot

By Simon Yoon

Canonical URL: https://www.tokenpost.com/news/technology/59753
Published: 2026-11-17T18:46:16.000Z
Updated: 2026-11-17T18:46:16.000Z
Section: Technology

> The Flash model is in a limited-access Microsoft Foundry preview. Its reported AT&T test results have not been confirmed by independent benchmarks.

Microsoft announced Nov. 17 that MAI Code 2 and MAI Code 2 Flash are headed to GitHub Copilot, adding two in-house coding models without giving a rollout date.

MAI Code 2 Flash is also available in a limited-access preview through Microsoft Foundry. Microsoft has not detailed broader availability beyond that preview.

In AT&T evaluations, MAI Code 2 Flash outperformed certain GPT-6.1 variants at 50 times lower cost. The comparison covered specific variants, not the GPT-6.1 family as a whole, and independent benchmarks have not verified the performance claims.

The models follow MAI Code 1 Flash, introduced in June, and MAI Code 1.1 Flash, released in August with native image understanding and pricing about 73% below its predecessor. Microsoft’s [Ignite expansion](<https://www.tokenpost.com/news/technology/59743>) included models for cybersecurity and other uses.



## Links in this article

- [Ignite expansion](https://www.tokenpost.com/news/technology/59743)
