GitHub Copilot adds OpenAI's GPT-5.6 Sol, Terra, and Luna across its coding surfaces

GitHub is rolling OpenAI's GPT-5.6 family into GitHub Copilot, adding Sol, Terra, and Luna as selectable models across the company's coding surfaces. The change matters because it is not limited to one IDE plugin or preview sandbox. GitHub says the rollout reaches Visual Studio Code, Visual Studio, Copilot CLI, the Copilot cloud agent, the GitHub Copilot app, github.com, GitHub Mobile on iOS and Android, JetBrains, Xcode, and Eclipse.

GitHub's official changelog confirms how the new lineup is meant to work inside Copilot. Sol is positioned as the highest-end option for complex reasoning over large codebases and long-running agentic tasks. Terra is the balanced default for everyday interactive and agentic coding. Luna is the lightweight, lower-cost option for smaller and faster jobs. GitHub also says the models use provider list pricing under Usage Based Billing, and that enterprise administrators must explicitly enable the GPT-5.6 policy because it is off by default for Copilot Business and Enterprise tenants.

The story is trending on X because it combines two things developers on the platform care about right now: model choice and where that choice actually shows up in real workflows. GitHub's own @GHchangelog post pushed the rollout directly into the X conversation, while follow-on discussion has centered on whether Copilot users should treat Sol as the premium reasoning pick, Terra as the new daily driver, or Luna as the practical cost-control option. That is a more concrete debate than the usual benchmark cycle, because the announcement lands inside tools developers already use every day rather than in a standalone API release.

For developers, builders, and product teams, the practical takeaway is that GitHub is making Copilot feel more like a model-routing layer than a single-assistant product. Teams can now choose a high-reasoning model for harder refactors and agentic work, a balanced model for normal coding sessions, or a cheaper model for lightweight tasks without leaving the same Copilot surfaces. That matters even more now that model access and request economics are becoming part of normal engineering decisions instead of hidden infrastructure details.

What remains unclear is how different the real-world behavior of Sol, Terra, and Luna will feel once the rollout reaches more accounts. GitHub says availability is gradual, so not every user will see the new models immediately. The company also has not answered the bigger workflow question yet: whether teams will develop stable habits around one default model, or start routing tasks much more deliberately based on codebase size, latency tolerance, and budget.