Ignite 2026, day one
● Generally available today MAI-Code-2 and MAI-Code-2-Flash, both in GitHub. MAI-Code-2-Flash is also in private preview in Microsoft Foundry.
● In preview MAI-Image-2.7, MAI-Image-2.7-Flash, and MAI-Cyber-2-Flash, the last inside Codename MDASH.
● Also generally available Microsoft-Decision-1.1, in Microsoft Foundry, OpenRouter and Vercel.
● The number nobody published A price. Microsoft's Book of News carries no per-token cost for any model announced.
Microsoft opened Ignite 2026 on 17 November in San Francisco with six new models, two of them generally available the same day. The conference runs to 20 November.
What each model is for
MAI-Code-2 is the one aimed at demanding agentic coding work, and it is generally available in GitHub now. MAI-Code-2-Flash handles high-volume everyday coding, is generally available in GitHub, and is additionally in private preview in Microsoft Foundry.
MAI-Image-2.7 and MAI-Image-2.7-Flash are image generation models with what Microsoft describes as sharper editing and more consistent output. Both are in preview.
MAI-Cyber-2-Flash covers vulnerability discovery, validation and remediation. It is preview only, inside Codename MDASH. Microsoft notes that MAI Cyber 1.1 Flash is the currently supported version there and that MAI-Cyber-2 will be used after evaluation testing - so the newest cyber model is not the one doing the work yet.
Microsoft-Decision-1.1 is the odd one out, and the most interesting. It is covered in a separate piece, because what it is built on is a story of its own.
Six, not seven
Several outlets have reported seven new models. Microsoft's own Book of News lists six: four MAI models, one MAI cyber model, and Microsoft-Decision-1.1, which does not carry the MAI prefix. If you are counting MAI-prefixed models only, the answer is five.
The pricing gap
The framing around this launch is that Microsoft is undercutting Anthropic and OpenAI on cost. That may turn out to be true. It is not something Microsoft has published. There is no per-million-token figure for MAI-Code-2, MAI-Code-2-Flash, or any other model in the announcement.
For MAI-Code-2 and its Flash variant, access runs through GitHub, so what you actually pay is likely to be governed by your GitHub Copilot arrangement rather than by a published token price. Until Microsoft says so, treat any cost comparison against Claude or GPT-6 as an estimate rather than a fact. We have compared the published coding-model prices elsewhere, and Microsoft is not in that table because it has not given us a number to put in it.
What else shipped alongside
Copilot Code enters the Frontier program now and reaches general availability in December. Microsoft IQ and Web IQ are generally available, Web IQ to all Azure customers. Copilot SDK 2.0 is generally available. Azure Cobalt 200 VMs are generally available and deploying to 24 Azure regions by year end, and Azure HorizonDB is generally available with Fabric HorizonDB in preview.
FAQ
Is MAI-Code-2 available now?
Yes. MAI-Code-2 and MAI-Code-2-Flash are both generally available in GitHub as of Ignite 2026. MAI-Code-2-Flash is additionally in private preview in Microsoft Foundry.
How much does MAI-Code-2 cost?
Microsoft has not published a price. The Ignite Book of News gives no per-token cost for any model in this announcement. Access runs through GitHub, so your Copilot arrangement is the practical answer until Microsoft publishes one.
Is MAI-Cyber-2 usable yet?
Not in production. MAI-Cyber-2-Flash is preview only inside Codename MDASH, and Microsoft says MAI Cyber 1.1 Flash remains the supported version there, with MAI-Cyber-2 to be used after evaluation testing.
Were there six models or seven?
Microsoft's Book of News lists six: MAI-Code-2, MAI-Code-2-Flash, MAI-Image-2.7, MAI-Image-2.7-Flash, MAI-Cyber-2-Flash and Microsoft-Decision-1.1. Five of those carry the MAI prefix.