At its November 17, 2026, Ignite announcements, Microsoft positioned MAI Code 2 for complex coding and MAI Code 2 Flash for routine tasks. The company also reported benchmark-specific performance and price comparisons for the models; those figures are tied to particular coding benchmarks, not every kind of AI workload. Microsoft’s Code 2 announcement
Match the model to the work: Microsoft describes Code 2 for complex, multi-file coding and Code 2 Flash for frequent, routine edits.
What each coding model is designed to handle
Microsoft describes MAI Code 2 as a model for deep debugging, multi-file implementation, reasoning across a code repository and user-interface work. The company says it is not yet tuned for fundamental science research or very long-horizon agentic work.
MAI Code 2 Flash is aimed at high-volume coding tasks such as explaining code, making targeted edits, fixing routine issues and iterating quickly. That makes the stated distinction practical: Code 2 targets work that calls for reasoning across a larger codebase, while Flash is positioned for repeated, narrower tasks. These are Microsoft’s descriptions of the models’ intended uses. Microsoft’s model details
What Microsoft’s benchmark comparisons measure
Microsoft reports Code 2’s average performance across four benchmarks—DeepSWE 1.1, Terminal Bench 2.1, SWE-Atlas and FrontierCode v1.1—and compares its price with named models. For Code 2 Flash, Microsoft reports comparable performance across three benchmarks for routine coding tasks. The company’s figures are:
| Microsoft model | Comparator | Reported performance relationship | Price as a share of comparator’s price | Benchmark scope |
| MAI Code 2 | GPT-6.1 Sol | 98% of its performance | 40% | Average across four benchmarks aimed at everyday enterprise coding |
| MAI Code 2 | Claude Opus 5.5 | 96% of its performance | 5% | Average across four benchmarks aimed at everyday enterprise coding |
| MAI Code 2 Flash | GPT-6 Luna | Comparable performance | 65% | Three benchmarks for routine, high-volume coding |
| MAI Code 2 Flash | Claude Haiku 5.5 | Comparable performance | 20% | Three benchmarks for routine, high-volume coding |
These are Microsoft-reported comparisons, not independent benchmark results. The percentages compare each model with a named competitor under the stated benchmark scope; they do not give an absolute per-token price for Code 2 or establish how either model will perform on a particular team’s code. Microsoft’s benchmark and price comparisons
Where Microsoft said the models were available
Microsoft said Code 2 and Code 2 Flash were generally available in GitHub Copilot. Its model-lineup announcement described Code 2 Flash in Microsoft Foundry as a limited-access preview. Microsoft’s Ignite announcement Microsoft AI’s model lineup
Microsoft also presented Microsoft IQ as an organizational-context platform. The company says it connects information across Fabric, Foundry and Microsoft 365 with third-party context, then turns that context into business plugins that can be shared across an organization. That is the broader context layer Microsoft announced alongside its task-specific models. Microsoft’s Ignite keynote