What GitHub says Terra is for
GitHub describes Terra as built for everyday agentic coding and terminal work at a lower cost than GPT-6.1 Sol. That makes it the mid-tier option. Sol is the "powerful" model and Terra is the one you leave running on routine jobs.
GitHub also makes a performance claim. In early testing, Terra solved noticeably more terminal and C# coding tasks than GPT-5.6 Terra, and it works in focused, consolidated steps on well-scoped tasks.
Treat that as a vendor statement. The announcement gives no task counts, scores or methodology. A third-party write-up of the launch made the same point, saying GitHub did not publish the underlying task counts or scores. "Noticeably more" is a direction, not a measurement.
"Well-scoped tasks" is also a hint about where Terra fits. This is a model for clear jobs like fixing a failing test, scripting a shell workflow or making a bounded refactor. Open-ended architectural work is a different case, and GitHub's wording does not claim Terra handles it.
Who can use it, and where
Terra is available to Copilot Pro, Pro+, Max, Business and Enterprise users. Copilot Free and Student plans are not on that list. GitHub's supported-models documentation says those plans get models through auto model selection only.
You can pick it in the model picker in these places:
- Visual Studio Code
- Visual Studio
- Copilot CLI
- GitHub Copilot cloud agent
- GitHub Copilot app
- github.com
- GitHub Mobile on iOS and Android
- JetBrains IDEs
- Xcode
- Eclipse
That covers the usual Windows developer stack, including both Visual Studio Code and full Visual Studio. GitHub's announcement does not give a minimum client or extension version for Terra. GitHub's supported-models documentation lists minimum versions for several other models, but not this one. Keeping your IDE and Copilot extension current is sensible general advice, though it isn't a Terra-specific requirement.
Why you might not see it yet
"Generally available" does not mean everyone sees it today. GitHub says the rollout will be gradual and asks users who don't see it to check back soon. No completion date was given.
If you are on Business or Enterprise, there is a second possible cause, which is organizational policy. A missing model on a paid plan isn't proof that your plan is ineligible.
Here is a sensible order of checks:
- Wait and refresh. Restart the IDE or reload the model picker. A staged rollout is the most likely explanation on launch day.
- Update your client. This is general hygiene and not a documented requirement for Terra.
- Check your plan. Free and Student plans aren't in the eligible list.
- Ask your admin if you are on Business or Enterprise.
What administrators need to know
Copilot Enterprise and Business administrators can manage access to Terra through the model policy in Copilot settings. The default behavior matters most. Under default model enablement, new models are enabled automatically unless an administrator has turned off the global default or explicitly disabled that model.
Two situations follow from that:
- Default left on: Terra will appear as the rollout reaches your organization, with no action from you. If you have compliance or cost-control concerns, disable it explicitly now.
- Global default turned off: Terra stays hidden until you enable it in its model policy.
GitHub's documentation adds that models you haven't explicitly configured show a "Delegate to Default Policy" label and inherit the default until you configure them. The same page lists categories that stay disabled by default regardless of that setting, such as pre-GA models and open-weight models. Terra is announced as GA, so those carve-outs shouldn't apply to it.
Billing: "provider list pricing" is not a price
GitHub says Terra is billed at provider list pricing under usage-based billing. The announcement gives no per-token figures.
GitHub's pricing documentation explains the mechanics. Interactions consume input, output and cached tokens. The total is converted into AI credits, where one credit equals $0.01. Individual plans include credit allowances. Business and Enterprise allowances are pooled at the billing-entity level. Usage beyond the allowance is billed at per-token rates.
At the time of writing, the retrieved GitHub pricing table did not include a GPT-6.1 Terra row. It listed GPT-6.1 Sol at $2.00 per million input tokens and $10.00 per million output tokens at the default tier. It listed GPT-5.6 Terra at $2.00 input and $12.00 output. So "lower cost than Sol" is GitHub's positioning, and the table I could check doesn't show it.
Some outside sites have published figures. One, DataStudios, reports OpenAI API pricing of $1 per million input tokens and $5 per million output tokens for Terra, priced at half of GPT-6.1 Sol in the API. That is a secondary report. I could not confirm it against GitHub's own pricing page, and Copilot's rates could differ. Check the live pricing page before you budget around it.
Two practical cost notes from GitHub's documentation:
- Larger context windows and higher reasoning levels consume more tokens and therefore more credits. GitHub recommends the regular settings by default.
- Code completions and next edit suggestions are not billed in AI credits.
How Terra fits the model lineup
Copilot's model list is changing quickly. On October 15 GitHub announced that GPT-5.6 Terra will be deprecated on November 16, 2026, with GPT-6.1 Sol or GPT-6 Luna suggested as alternatives. I mean that as a separate notice. GitHub's announcement does not say Terra 6.1 replaces 5.6 Terra, and that deprecation notice came before this release and names other alternatives. Teams with workflows pinned to GPT-5.6 Terra have until mid-November to pick a new default. Terra 6.1 is now an obvious candidate to test, but nothing in GitHub's notices makes it the official successor.
Practical advice
- Test on your own repos. GitHub's C# and terminal claim is plausible and untested by outsiders. Run it on a few real tickets and compare against your current default.
- Watch credit consumption. Cheaper per token doesn't always mean cheaper per task, since a model that takes more steps or longer output can erase the gap. That is general reasoning, not a Terra finding.
- Keep reviewing the output. GitHub's documentation tells users to review and validate generated code, including for security, with thorough human review before production use. That is general Copilot guidance, not a Terra-specific warning, but it applies to agent-mode work in a terminal.
- Set policy deliberately if you administer a tenant. Default enablement is convenient, but it means new models turn on automatically.
Bottom line
For developers, the news is a new mid-priced model option across nearly every major Copilot surface. The evidence for it is GitHub's qualitative early testing, with no published benchmark. For admins, the news is that Terra will switch on by default unless you have configured otherwise. For anyone budgeting, the exact Terra rate is not in the pricing table I could check, so confirm it in your billing documentation before moving production workloads.