Microsoft coding model · access policy rechecked August 28, 2026

MAI-Code-1.1-Flash balances coding quality, speed and cost.

It adds native vision and is rolling out across GitHub Copilot. Availability still depends on plan, client, rollout and policy; vendor benchmarks still need a workload test.

Verified model record

Provider

Microsoft AI; generally available in GitHub Copilot during the rollout announced August 11, 2026.

Best fit

Lightweight agentic coding, everyday engineering, CLI, .NET, tool use, and screenshot-to-prototype workflows that pass your own evaluation.

Inputs

Text plus native vision for understanding screenshots, diagrams, and designs; verify support on the exact Copilot surface.

GitHub provider price

$0.20 input, $0.02 cached input, and $1.20 output per million tokens as currently listed; provider pricing can change.

Copilot request treatment

GitHub announced a 0.25× premium request multiplier for annual subscribers; other plans and usage-based billing require a current check.

Business and Enterprise access

An explicit model setting wins. If the GA model is unconfigured, it can inherit the current Default availability for released models policy; exclusions, plan, client, rollout, data-residency, and FedRAMP rules still apply.

Flowith access

Not asserted. Microsoft and GitHub availability does not establish a Flowith workspace model mapping.

Provider evidence, not a guarantee

Microsoft reports 22% better Terminal-Bench 2.1 results in Copilot CLI, 15% better .NET results, 4% higher code survival, 9% more return visits, 25% faster streaming and 25% fewer tokens versus MAI-Code-1-Flash. These are vendor benchmarks and production telemetry; test your repositories, languages, tools and review burden.

Access is a decision tree

Free and Student users receive it through automatic selection. Paid Copilot tiers can manually select it after rollout. For Business and Enterprise, an explicit model setting wins; an unconfigured GA model can inherit the current default-availability policy. Exclusions and live plan, client, rollout, data-residency, and FedRAMP constraints remain authoritative.

Continue by decision

Vision and costTest inputs, quality, speed and billing.

Official sources

Microsoft model page — capability, vision, Copilot focus and model-card links.
Microsoft launch evidence — vendor benchmarks, production telemetry, speed, token efficiency and relative cost.
GitHub Copilot availability — plans, surfaces, rollout, request treatment and policy.
GitHub default availability — unconfigured GA inheritance, explicit settings, exclusions, and administrator controls.
GitHub supported models — current release status, plan, client, and default-eligibility reference.

MAI-Code-1.1-Flash questions

It is Microsoft's lightweight agentic coding model for everyday engineering workflows, now generally available through GitHub Copilot during its announced rollout.

Yes. Microsoft and GitHub describe native vision for understanding screenshots, diagrams, and designs. Verify image input on the exact Copilot client and version.

GitHub currently lists provider pricing of $0.20 input, $0.02 cached input, and $1.20 output per million tokens. Copilot request accounting is a separate plan-specific layer.

No. GitHub's launch states 0.25x for annual subscribers. Current plan, billing model, included allowance, and usage-based pricing must be checked separately.

Not necessarily. An explicit enable or disable setting wins; otherwise this GA model can delegate to the current default-availability policy. Plan, client, rollout, data-residency, FedRAMP, and live-account constraints still apply.

This page does not claim a Flowith mapping. Check the current Flowith model selector separately from Microsoft and GitHub availability.