GitHub has added Google’s Gemini 4 Argon to GitHub Copilot, giving eligible subscribers another model for difficult programming work and tasks that require reasoning. Access covers Copilot Pro+, Max, Business and Enterprise subscribers, with charges calculated from the provider’s list prices through usage-based billing. The model has reached general availability, though GitHub is distributing access gradually, so some eligible users may still be waiting for it to appear.
Users can choose Gemini 4 Argon through the model picker in tools including Copilot CLI, Visual Studio Code and JetBrains IDEs. GitHub also supports the model in its GitHub Copilot app and GitHub Copilot coding agent, extending the choice beyond editor-based conversations. For organizations on Business or Enterprise plans, administrators can control whether users get access through the model policy within Copilot settings.
Read more: GitHub Adds Claude Haiku 5.5 to Copilot for Quick Coding Tasks
GitHub says Gemini 4 Argon completed more tasks than earlier Gemini models across every coding benchmark used in its early tests. The company reports that the biggest improvements came in Copilot CLI. GitHub also says the model tests edge cases and checks its work before completing a task, and describes it as intended for complex coding, reasoning and agentic assignments.
Under GitHub’s default model enablement rules, newly introduced models become available automatically unless an administrator has switched off the global default or disabled that particular model. Separately, GitHub introduced controls for JetBrains on October 10 that let Enterprise administrators select an available Copilot agent model as the default when users start conversations. Earlier this month, GitHub deprecated Gemini 3.5 Flash and Gemini 3.6 Flash across Copilot experiences, recommending Gemini 3.8 Flash as the alternative for both.
The Copilot addition arrives the same day as Google’s October 13 expansion of Gemini 4 Argon to the Gemini app for Google AI Pro and Ultra subscribers, alongside the ability to assign Gemini tasks. Google’s initial October 1 rollout had targeted trusted cyber defenders through the Fairwind Program, with broader audiences promised but no access timetable given at that stage. That earlier introduction listed introductory token rates of $2 per million for input and $10 per million for output, separate from GitHub’s current statement that Copilot billing uses provider list pricing.
