GitHub Copilot: MAI-Code-1-Flash Reaches Business and Enterprise Plans

MAI-Code-1-Flash Expands to GitHub Copilot Business and Enterprise

GitHub has brought MAI-Code-1-Flash, Microsoft AI's first in-house coding model, to Copilot Business and Copilot Enterprise plans. The June 26 update completes the rollout of the model across all major Copilot subscription tiers, following its initial launch at Microsoft Build in early June and its expansion to eight Copilot surfaces on June 18 for individual plan users.

What Is MAI-Code-1-Flash?

MAI-Code-1-Flash is a coding model built end-to-end by Microsoft AI for the GitHub Copilot and VS Code tooling stack. It uses a sparse Mixture-of-Experts (MoE) architecture with only 5 billion active parameters per inference token, delivering token efficiency approaching that of a 70B dense model while remaining significantly faster and cheaper to run.

The model was trained without distillation from third-party models, relying on what Microsoft describes as clean, traceable, and enterprise-grade training data. In Microsoft's benchmark evaluations, MAI-Code-1-Flash scored 51.2% on SWE-Bench Pro versus 35.2% for Anthropic's Claude Haiku 4.5, and used up to 60% fewer tokens on SWE-Bench Verified.

Where It Fits in the Copilot Lineup

MAI-Code-1-Flash is positioned for high-volume, iterative tasks rather than the most demanding architectural or deep-debugging challenges. It excels at fast code completions, inline suggestions, chat-based explanations, and agentic coding loops where low latency and cost efficiency matter more than maximum reasoning depth. Developers who rely on a larger model for complex design decisions can still route the frequent, smaller steps of a working session to MAI-Code-1-Flash.

How to Enable It on Business and Enterprise Plans

Unlike the rollout to individual tiers, access to MAI-Code-1-Flash on Business and Enterprise plans is not automatic. Administrators must navigate to Copilot settings and explicitly enable the MAI-Code-1-Flash policy before any developers in the organization can select or be routed to the model. This opt-in approach gives organizations control over which models their teams use, consistent with the existing policy management framework for enterprise Copilot.

Pricing

MAI-Code-1-Flash is billed under GitHub's usage-based billing system at provider list pricing: $0.75 per million input tokens, $0.075 per million cached input tokens, and $4.50 per million output tokens.

What This Means for Teams

For development teams on Business and Enterprise plans, this update opens up a fast, low-cost option for the high-frequency parts of AI-assisted workflows. Teams running large volumes of agentic coding tasks, generating tests, doing repetitive refactoring, iterating on implementation details, can route those requests to MAI-Code-1-Flash and reserve more powerful models for work that genuinely benefits from deeper reasoning.