Microsoft rolls out MAI‑Code‑1‑Flash

Microsoft introduced MAI-Code-1-Flash on June 2 as a new in-house coding model for GitHub Copilot, and said it is rolling out to individual users in Visual Studio Code. The company said the model was built “end-to-end by Microsoft” using “clean and appropriately licensed data” and was designed for fast coding help in day-to-day developer workflows. GitHub, which Microsoft owns, said availability is beginning with a limited set of users and will expand over the coming weeks. Microsoft positioned the release as part of a broader push to put more of its own models into developer products. ### Where does this actually show up for developers? Visual Studio Code is the first place Microsoft is putting the model in front of Copilot users. Microsoft said MAI-Code-1-Flash appears in the model picker and also under the default auto picker for GitHub Copilot individual users in VS Code. GitHub’s changelog said the rollout covers Copilot Free, Pro, Pro+ and Max plans, though it starts with a limited group before widening. (microsoft.ai) GitHub’s wording matters because it frames the release as a gradual product rollout, not a universal switch. Developers who do not see the model immediately may get access later as availability expands over the next several weeks, according to the changelog. ### What did Microsoft say the model is built to do? Microsoft said MAI-Code-1-Flash is aimed at “fast, efficient assistance in everyday developer workflows.” On its product page, the company said the model handles agentic coding in real developer environments, follows instructions across single-turn and multi-turn tasks, and uses “adaptive thinking” so it stays brief on simple requests while spending more reasoning effort on harder ones. (microsoft.ai) (github.blog) The company also said it built the model around production workflows rather than benchmark performance alone. Microsoft said that choice was meant to make the model work better inside the GitHub Copilot harness developers already use. ### How is Microsoft describing its performance? Microsoft said MAI-Code-1-Flash “outperforms Claude Haiku 4.5 with better price to performance across coding benchmarks.” The company did not provide the full benchmark table in the GitHub changelog entry, but it used that comparison in its product announcement to argue the model is designed to deliver coding help with lower cost and higher efficiency. (microsoft.ai) CNBC reported on June 2 that MAI-Code-1-Flash is Microsoft’s first model in the AI coding category and part of a larger effort to compete more directly with OpenAI, Anthropic and Google using proprietary models. That characterization came as Microsoft unveiled several new models at its Build developer conference in San Francisco. (microsoft.ai) ### Why is GitHub Copilot central to this release? GitHub said MAI-Code-1-Flash was “designed and tuned specifically for GitHub Copilot” and called it the first in a new wave of purpose-built coding models from Microsoft. Microsoft’s product page used similar language, saying the model was trained and designed for the GitHub Copilot harness so the two would work better together. (cnbc.com) That makes Copilot the launch vehicle as much as the product destination. Rather than releasing the model first as a general API, Microsoft is introducing it inside a coding assistant already embedded in a widely used editor. ### What happens next in the rollout? The next step is a broader Copilot rollout over the coming weeks. GitHub said MAI-Code-1-Flash is beginning with a limited set of users across Free, Pro, Pro+ and Max plans, and developers will be able to select it in the Visual Studio Code model picker as access expands. (microsoft.ai) Microsoft’s June 2 product page remains the main source for model details, while GitHub’s June 2 changelog tracks availability inside Copilot.