GitHub Adds Gemini 3.8 Flash to Copilot Across Developer Tools

Developers using GitHub Copilot are gaining a new model choice for coding tasks across their editors, command line and cloud-based agent workflows. GitHub has added Gemini 3.8 Flash to Copilot for paid individual and organizational plans, though the rollout is gradual rather than universal on day one.

The practical change is model selection inside tools developers already use. Eligible Copilot Pro, Pro+, Max, Business and Enterprise users can choose Gemini 3.8 Flash from the model picker as it becomes available to their accounts. GitHub says the model is rolling out across Visual Studio Code, Visual Studio, Copilot CLI, the Copilot cloud agent, the Copilot app, JetBrains IDEs, Xcode and Eclipse.

That breadth puts the same Copilot model option in interactive coding surfaces and tools built for longer-running work. A developer may encounter it in an editor or command line, while teams using Copilot’s cloud agent will see it in that environment too. GitHub has not given a timetable for when every eligible user will receive the option.

A model choice, not a blanket enablement

For organizations, availability is not solely a developer-level decision. Copilot Business and Enterprise administrators can manage access to Gemini 3.8 Flash through the product’s model policies. Under GitHub’s default model-enablement setting, new models are enabled automatically unless an administrator has switched off the global default or explicitly disabled this model.

GitHub’s coding assessment has a narrow scope

GitHub said its early testing found Gemini 3.8 Flash performed strongly on complex terminal-based coding tasks. It also said the model showed rigorous validation and persistent recovery from actionable failures—useful behaviors for work where a coding system needs to check results and continue after an error rather than stop at its first attempt.

Those are GitHub’s early-testing observations, not a published benchmark comparison. The changelog provides neither scores nor a test design, so users evaluating the model will need to judge its performance in their own repositories and workflows. For now, the concrete decision points are simpler: whether the staged rollout has reached their environment, whether an organization’s policy permits the model, and how its introductory usage-based pricing fits their work.