AI Models & Platforms
Microsoft AI Releases MAI Code 2 and MAI Code 2 Flash for Enterprises

Key PointsMicrosoft positions MAI Code 2 on cost per task rather than raw benchmark superiority…
View more
- Microsoft positions MAI Code 2 on cost per task rather than raw benchmark superiority over rival models.
- The Flash variant beats both comparators on average pass rate while reporting a lower cost per task.
- A Flash derivative moves into Microsoft 365 and Fabric IQ extending the release beyond coding into everyday enterprise work.
Microsoft AI announced two models for everyday enterprise coding, MAI Code 2 and MAI Code 2 Flash, on November 17, 2026, reporting that MAI Code 2 delivers 98% of GPT-6.1 Sol’s performance at 40% of its cost and 96% of Claude Opus 5.5’s performance at 5% of its cost across four benchmarks.
MAI Code 2 is aimed at production changes, deep debugging, multi-file implementation, repository-wide reasoning and user interfaces. The Flash variant targets high-volume jobs: explaining code, making targeted edits, applying routine fixes and iterating quickly. The split separates deeper implementation work from frequent tasks where lower per-task costs matter at scale.
Microsoft is explicit about the boundary: neither model has yet been tuned for fundamental science research or very long-horizon agentic work. That limitation defines the release’s scope around the coding work customers do every day.
Benchmark Results and Cost Conditions
Across DeepSWE 1.1, Terminal-Bench 2.1, SWE Atlas and FrontierCode v1.1, Microsoft reports a 63.2% average pass rate and $0.309 average cost per task for MAI Code 2. GPT-6.1 Sol averaged 64.1% at $0.761, while Claude Opus 5.5 averaged 65.3% at $6.306. The small pass-rate gaps and much larger cost differences explain the release’s price-performance positioning.
All models in the coding comparisons ran at maximum reasoning effort. The cost figures for Microsoft’s models use a 50% promotional price, and the company cautions that changing the workload, model configuration, reasoning effort or pricing assumptions can change the comparisons. DeepSWE, Terminal-Bench and SWE Atlas ran on Microsoft’s internal MSBench platform through GitHub Copilot CLI without subagents.
For MAI Code 2 Flash, Microsoft reports a 61.0% average pass rate and $0.052 average cost per task across DeepSWE 1.1, Terminal-Bench 2.1 and SWE Atlas. GPT-6 Luna averaged 55.1% at $0.080, and Claude Haiku 5.5 averaged 53.9% at $0.319. Flash therefore has a higher average pass rate and lower reported task cost than either comparator.
Separate blind frontend evaluations covered user interfaces, games, data visualizations, 3D scenes and SVG graphics. Microsoft reports that 258 annotators preferred MAI Code 2 over GPT-6 Sol 67% of the time on aesthetics and functionality.
Copilot Availability and Product Integration
Both models are available in GitHub Copilot. MAI Code 2 Flash is also in limited-access preview in Microsoft Foundry, with releases planned for OpenRouter, Vercel, Hermes, OpenCode and Kilo.
Microsoft has further optimized Code 2 Flash into MAI Work Flash for use across Microsoft 365 and Fabric IQ. The company reports comparable accuracy and token efficiency with GPT-6 Luna on internal Excel customer-scenario benchmarks, and similar quality and token efficiency on frequent PowerPoint tasks.
In Teams evaluations, Microsoft reports higher scores than GPT-6 Luna for decision tracking and knowledge gaps, adapting custom instructions and tool calling, with up to 28% fewer tokens on instruction following and tool calling.
For Power BI semantic models, Microsoft reports a 0.77 overall score for MAI Work Flash in offline DAX query-generation evaluations, compared with 0.54 for GPT-6 Astra, 0.45 for GPT-5.6 Sol and 0.30 for Opus 5.5. DAX query generation is the capability behind Power BI’s Chat with your Data feature.
Microsoft’s model release is therefore also a distribution release: Copilot gets two coding options, while a Flash derivative moves into spreadsheets, presentations, chat and analytics. The reported numbers make cost per task the main competitive lever, with promotional pricing and the deliberately narrow workload focus determining how far the comparisons extend.












