MAI-Code-1-Flash Is Now Generally Available for Copilot Business and Enterprise

MAI-Code-1-Flash, Microsoft’s in-house coding model, is now generally available for GitHub Copilot Business and Copilot Enterprise customers. The update posted to the GitHub Changelog on June 26.

What Changes for Organizations

MAI-Code-1-Flash was already available to individual Copilot tiers through the VS Code model picker and Copilot’s Auto router. The June 26 GA extends that to Business and Enterprise plans, which is where most teams actually run.

There’s one catch: administrators have to explicitly enable the model before anyone can use it. Enterprise and Business plan admins need to turn on the MAI-Code-1-Flash policy in Copilot settings. It doesn’t roll out automatically.

Once enabled, the model is available in the same places as other Copilot models: VS Code, Copilot Chat, Copilot CLI, cloud agents, and GitHub Mobile.

The Model

MAI-Code-1-Flash is a 5-billion-parameter model trained specifically for coding workflows inside GitHub Copilot. Microsoft says it was trained “from the ground up on clean, traceable and enterprise-grade data, without distillation from third-party models,” which is the company’s way of distinguishing it from fine-tuned versions of models from other labs.

On SWE-Bench Pro, Microsoft benchmarks it at 51.2%, compared to 35.2% for Claude Haiku 4.5. It’s also described as solving complex tasks with up to 60% fewer tokens, which is the more practical metric for organizations paying attention to usage costs under Copilot’s AI Credits billing model.

The model uses provider list pricing under usage-based billing, meaning it draws from the same credit pool as other Copilot models rather than having a separate quota.

When It Makes Sense

MAI-Code-1-Flash is positioned for high-volume, iterative agentic coding workflows where speed and latency matter more than maximum capability. It’s not trying to compete with Copilot’s larger models on complex tasks. The use case is closer to autocomplete and fast code review at scale, where burning a large model on every call would run through AI Credits quickly.

For organizations already on Copilot Business or Enterprise, it’s worth enabling to see whether the latency difference is perceptible in daily use before deciding whether the benchmark numbers translate to your actual codebase.

---

Sources: GitHub Changelog – MAI-Code-1-Flash GA / Microsoft AI – Introducing MAI-Code-1-Flash