GitHub Copilot adds Gemini 4 Argon, Google's new model
IA en un minuto newsroom · Editor: Jon Elgezabal
In 30 seconds
Gemini 4 Argon, Google's model built for coding and reasoning, can now be chosen in GitHub Copilot on the Pro+, Max, Business and Enterprise plans. GitHub says that in its early trials it beat earlier Gemini models on every coding test, especially in Copilot CLI. It is charged by usage at Google's list prices and is arriving gradually, so it may take a while to show up.
GitHub announced on Tuesday, October 13, that Gemini 4 Argon, Google's latest frontier model, can now be used by everyone eligible in GitHub Copilot, its AI coding assistant. It is open to people on the Copilot Pro+, Max, Business or Enterprise plans. The release is staggered, so some users may not see it yet.
GitHub describes it as a model for demanding programming work, agent tasks (where the AI carries out several steps on its own) and reasoning. According to the company, in its first tests Gemini 4 Argon completed more tasks than earlier Gemini models across all the coding benchmarks it ran, and its biggest improvement came in Copilot CLI, the version of Copilot used from the terminal. GitHub adds that the model reviews its own work carefully and tries out edge cases before it wraps up. The announcement gives no figures for those tests.
It appears in the model selector of Visual Studio Code, Visual Studio, JetBrains development environments, Xcode, Eclipse, Copilot CLI, the Copilot coding agent and the GitHub Copilot app. Use is charged by consumption, at the list prices set by Google, the model's provider.
On the Business and Enterprise plans, administrators decide who gets the model through the model policy in Copilot's settings. With the default setting, new models are switched on automatically unless the administrator has turned that general option off or blocks this particular model.
Google unveiled Gemini 4 Argon on September 30 and first gave it to a group of trusted cybersecurity defenders through its Fairwind program. On October 13 it widened access to Google AI Pro and Ultra subscribers in the Gemini app and to developers who pay for the Gemini API, through Google AI Studio and Android Studio; AI Ultra subscribers also have it in AI Studio and in Antigravity, its agent-based development platform.
Why it matters · analysis and opinion
Copilot's model picker now works as a shop window for several makers, and GitHub adding Google's model on the same day Google opened it to its paying customers shows how keen the industry is not to fall behind. For a development team it is one more option to try on long tasks, not a forced change. The tricky part lies in two details: it is charged by consumption, so it makes sense to watch spending before using it every day, and on company accounts it arrives switched on unless someone has closed it off. Anyone who administers Copilot in an organization should check their model policy now. And the improvements GitHub announces come without published data, so the comparison with other models is still to be made.
Official source: GitHub · Written with the help of AI: how we make the news
