Microsoft Brings OpenAI’s GPT-5.4 Agentic Coding Model to GitHub Copilot

OpenAI’s latest model has quickly found its way into Microsoft’s developer tools. Merely hours after OpenAI announcenced GPT-5.4 models, Microsoft rolled it out to GitHub Copilot. The company says that the new model brings improved reasoning and top-notch performance for real-world software development tasks.

Microsoft, in its announcement, describes the GPT-5.4 as OpenAI’s “latest agentic coding system,” which is designed to handle complex workflows that require multiple steps and tools. Per the company, early testing suggests GPT-5.4 performs better across a range of development scenarios, especially when dealing with intricate coding tasks.

GPT-5.4 rolls out across GitHub Copilot tools and platforms

Microsoft says GPT-5.4 will be available to Copilot Pro, Pro+, Business, and Enterprise users. Once enabled, developers can select the model directly through the Copilot model picker. Support spans a wide range of development environments. Developers will be able to use GPT-5.4 in Visual Studio Code version 1.104.1 and later across chat, ask, edit, and agent modes. Visual Studio version 17.14.19 and later also supports the model in agent and ask modes.

The rollout extends beyond Microsoft’s tools as well. JetBrains IDEs version 1.5.66 and later include support in ask, edit, and agent modes. Xcode version 0.48.0 and newer supports the model in ask and agent modes, while Eclipse version 0.15.1 and later also integrates GPT-5.4.
Developers using GitHub’s web platform and mobile apps are included too. The model will appear on GitHub.com, the GitHub CLI, and GitHub Mobile on both iOS and Android.
GitHub notes that while GPT-5.4 appears across these versions, the best results come when developers upgrade to the latest releases of each environment.

Microsoft further adds that administrators must first enable access by activating the GPT-5.4 policy in Copilot settings for organizations using Copilot Business or Enterprise plans. Once enabled, teams can begin testing the new model across supported tools.