
Microsoft launched MAI Code 2 and MAI Code 2 Flash in GitHub Copilot on November 17. Code 2 is built for hard jobs that span a codebase. Flash is meant for quick fixes, small edits and other tasks that come up each day.
Both models are now in the Copilot picker. Code 2 is open to Pro, Pro+, Max, Business and Enterprise users. Flash also comes with the Free and Student plans. Users on annual plans get access if their plan is on the list. For work accounts, admins can control access through a model policy. See GitHub’s list of plans and apps.
How MAI Code 2 and Flash differ
Microsoft trained Code 2 to work across a full codebase. It can help find a bug that spans more than one file. It can also help teams rework old code or build a new screen. Flash handles tasks with a smaller scope. A dev could ask it to explain a short block of code or change how a button works.
Microsoft ran Code 2 on four tests used to grade coding models. It says Code 2 reached 98 percent of GPT-6.1 Sol’s score at 40 percent of the cost. Both were set to their highest reasoning level. Three of the tests ran in Microsoft’s own test setup. The price used for Code 2 was half of its planned full rate. Microsoft’s launch post gives the test details.
What GitHub charges for each model
MAI Code 2 costs $0.50 per million input tokens, $0.025 for cached input and $2.50 for output. The listed Flash rates are $0.05, $0.005 and $0.25. These rates last until May 17, 2027. GitHub says both sets of rates will then double. One million output tokens will cost $5 for Code 2 or $0.50 for Flash.
GitHub bills model use in AI credits. Plans come with their own credit limits. A task that reads more code or needs more rounds of edits can use more tokens. Paid plans still get unlimited code completions and next edit suggestions. GitHub tracks those features on a separate count, outside the AI credit budget for model calls. GitHub’s pricing guide lists the rates and how billing works.
Where to find Code 2 and Flash
Users can pick the models in VS Code, Visual Studio, the Copilot app and the command line. They can also pick them in more apps, such as GitHub Mobile. GitHub added Code 2 to the Standard tier of Copilot Fusion. That mode can pick and team up models based on the effort level a user sets. In Microsoft Foundry, Flash is in a preview with limited access.
Microsoft first showed MAI-Code-1-Flash at Build earlier this year. GitHub will drop a later model in that line, MAI-Code-1.1-Flash, from Copilot on December 18. Teams that still use it have until then to switch their workflows. Admins who manage the list of models may need to turn on the new Flash model first. GitHub’s notice gives the date and steps to switch.
Read also
Popular news

OpenAI Launches GPT-6.2 Astra and Sol for ChatGPT Work, Codex and the API

We asked four AIs to predict the GTA 6 story. They all think Lucia and Jason survive

AI Acceptable-Use Policy Template for Small Businesses

Who Controls an AI Agent When Its Creator Leaves?

Nokia and Partners Plan a Four-Country AI Gigafactory Bid



