MAI-Code-1-Flash is now GA for Copilot Business and Copilot Enterprise

What's new

• MAI-Code-1-Flash GA for Business/Enterprise: available for Copilot Business and Enterprise admins, who must enable the policy.
• Optimized for agentic coding: 5B parameters, trained directly against the production Copilot harness.
• Performance: 16-point lead over Claude Haiku 4.5 on SWE-Bench Pro, solves same tasks with up to 60% fewer tokens.
• Billing: provider list pricing under usage-based billing.
• Policy controls: enterprise admins must enable MAI-Code-1-Flash in Copilot settings before users get access.

Why it matters

Microsoft is pushing MAI-Code-1-Flash as a cheaper, faster alternative to OpenAI/Anthropic models in the Copilot stack. For enterprise teams this means higher throughput at lower token cost, with full control via policy.

How to try

Copilot Business and Enterprise admins enable MAI-Code-1-Flash in Copilot settings → Models. Users then see it in the IDE model picker.