GitHub Copilot is adding three OpenAI GPT-5.6 models to its model picker: Sol, Terra, and Luna. That gives developers a choice between a model aimed at demanding reasoning, one positioned as the balanced everyday option, and one for smaller, faster tasks.
This is worth attention if your team already uses Copilot for coding work. It is not a reason to change your whole development process on announcement day.
What Changed
GitHub says GPT-5.6 is rolling out across Copilot. Sol is positioned for complex reasoning over large codebases and long-running agentic work. Terra is the balanced option for everyday interactive and agentic coding. Luna is the lightweight option for smaller, faster tasks.
Availability is not identical across plans. GitHub lists Sol for Copilot Pro+, Max, Business, and Enterprise. Terra and Luna are also listed for Copilot Pro. The models are intended to appear in the picker across products including VS Code, Visual Studio, Copilot CLI, the cloud agent, the Copilot app, GitHub.com, GitHub Mobile, JetBrains, Xcode, and Eclipse.
The rollout is gradual. If the models are not visible in your account yet, that does not mean the announcement is wrong. Business and Enterprise administrators also need to enable the GPT-5.6 policy, which is off by default.
Who Should Test It
Test this if you already use GitHub Copilot and have a repeatable task where the model choice could matter: understanding an unfamiliar part of a codebase, planning a multi-file change, or making a small, well-defined edit quickly.
This matters most for teams that can compare results with their current Copilot setup. If you do not have a stable task or a way to review the output, you will learn very little from switching models at random.
Where It Fits in a Workflow
Use the model picker at the point where you already ask Copilot for help. Try Sol when the task needs more context and careful reasoning. Try Terra as the normal comparison point. Try Luna when the task is small and speed matters more than a long chain of reasoning.
The model does not replace code review. Treat its output as a first pass: inspect the proposed changes, run the relevant checks, and keep the same human approval step you use today.
When to Skip It
Skip a broad rollout if the models are not yet available to the people who need them, if an administrator has not enabled access, or if usage-based billing is not understood by the team that owns the budget.
Also skip the experiment when the only test is a toy prompt. A model picker is useful when it changes work you already do—not when it produces a more impressive demo.
A 20-Minute Test
Pick one real task that has a clear review standard. Good examples: explain a module you did not write, plan a small multi-file change, or draft a focused test.
Run the task with the Copilot model your team normally uses. Save the prompt, time taken, and result.
Run the same task with Terra. If Sol or Luna is available and relevant, run one of those as a third comparison—not all three by default.
Review each result for correctness, useful context, time saved, and the amount of cleanup required.
Keep the new option only if it gives a clear benefit on your real task. If the answer is merely different, keep your current default.
Public Test Path
Open GitHub Copilot’s model picker in a supported product once the rollout reaches your account. For Business or Enterprise, ask the administrator to confirm that the GPT-5.6 policy is enabled first.