Qwen4-Max-Preview
A time-limited multimodal reasoning preview for interactive coding and agent tools through Alibaba Cloud Token Plan.
Coding model for debugging, multi-file implementation, repository reasoning, and interface development, available through eligible paid GitHub Copilot plans.
Visit website
MAI Code 2 is Microsoft's coding model for repository-based software engineering. It accepts text and images and produces text.
GitHub lists it as generally available in Copilot and recommends it for deep reasoning and complex, multistep coding. It is relevant to developers and engineering teams working on changes that require understanding relationships across a codebase.
Supports multi-file implementation, refactoring, debugging and test-driven verification.
Answers codebase questions and supports multistep shell, file and search workflows, depending on deployment.
Accepts screenshots and UI mockups for coding tasks.
Supports low, medium, high, xhigh and max reasoning effort.
Ask for an explanation of a failure and a proposed fix across the affected files.
Describe a structural change and request an implementation with relevant verification.
Provide a UI reference and ask for a corresponding implementation in an existing project.
Copilot usage consumes plan allowances, with metered token usage converted into GitHub AI Credits. Additional usage follows the plan's billing rules.
Promotional model rates until May 17, 2027; these are usage rates rather than subscription prices. Starting May 17, 2027, rates become USD 1 input, USD 0.05 cached input and USD 5 output per million tokens.
Yes. GitHub lists it as generally available, subject to plan, client and organization model policies.
No. Changing the Copilot Chat model does not change the model used for inline suggestions.
GitHub positions Flash for fast code completions, explanations and everyday coding. MAI Code 2 is positioned for deeper reasoning and complex multistep work.
Consider Flash for fast code completions and explanations; GitHub recommends MAI Code 2 for deeper reasoning and complex multistep coding.