MAI Code 2 and Code 2 Flash
- Source
- microsoft.ai
- Author
- lemonish97
- Date

Context
Microsoft has released two coding models, MAI Code 2 and MAI Code 2 Flash, and its headline claim is about cost. Averaged over four coding benchmarks, Microsoft says MAI Code 2 delivers 98% of the performance of OpenAI's GPT 6.1 Sol at 40% of the cost, and 96% of Claude Opus 5.5's performance at 5% of the cost. Code 2 is aimed at larger jobs such as multi-file changes, deep debugging and reasoning across a whole repository. Code 2 Flash is for high-volume work such as code explanations, targeted edits and routine fixes, where Microsoft says it performs comparably to GPT-6 Luna and Claude Haiku 5.5 at lower cost. Both models are available in GitHub Copilot, and Code 2 Flash is also in a limited-access preview in Microsoft Foundry, the company's platform for building on AI models.
Microsoft's own footnotes qualify those ratios in two ways. The MAI costs are calculated at a 50% promotional price, so the cost comparisons would be less favorable to Microsoft at the models' standard price. And Microsoft ran three of the four benchmarks itself, on its internal MSBench platform through the GitHub Copilot command-line tool, with every model at maximum reasoning effort and no subagents. The fourth, FrontierCode, uses scores and costs from Artificial Analysis, an independent benchmarking firm. Microsoft adds that results vary with workload, model configuration, reasoning effort and pricing assumptions.
Microsoft is also direct about where the models stop. It says they are not yet tuned for harder problems, naming fundamental science research and very long-horizon agentic work. The supported claim is therefore narrower than the headline: near-frontier results on everyday enterprise coding, measured mostly on Microsoft's own setup, at a discounted price. Copilot is a familiar route for these models; Microsoft's earlier MAI-Code-1.1-Flash rolled out there in August.
Checking sign-in…
Loading comments…


