GPT-6.1 Sol in GitHub Copilot: Availability, Cost and a Fair Test

Quick answer: GitHub announced GPT-6.1 Sol in GitHub Copilot on September 29, 2026. It is generally available as a model but rolling out gradually inside Copilot to Pro+, Max, Business and Enterprise users. GitHub lists Visual Studio Code, Visual Studio, Copilot CLI, the coding agent, its app, github.com, mobile and several other IDEs among the model-picker surfaces.

That announcement creates a practical search problem: readers want to know whether they have access, what they will pay and whether switching models will improve their own work. GitHub's early tests say the model used fewer tokens and steps than earlier GPT-6 and GPT-5.6 families on the tasks it tested. That is a vendor-reported observation, not a guarantee for your repository. A small local comparison is more useful than a generic ranking.

Where can I select GPT-6.1 Sol?

The GitHub release note lists Copilot Pro+, Max, Business and Enterprise. On a supported surface, open Copilot's model picker and look for GPT-6.1 Sol. GitHub names VS Code, Visual Studio, Copilot CLI, Copilot coding agent, the Copilot app, github.com, GitHub Mobile, JetBrains IDEs, Xcode and Eclipse. Rollout is gradual, so a missing entry on October 3 is not by itself proof that your subscription is ineligible.

Business and Enterprise administrators can control the model through Copilot's model policy. If a team member cannot see it, check the plan, the product's current version and the organization's policy before troubleshooting an individual machine. GitHub says default enablement can make new models available automatically unless an administrator disables that default or explicitly blocks this model.

How does billing work?

GitHub says GPT-6.1 Sol is billed at provider list pricing under usage-based billing. Copilot subscription access and model usage charges are different questions. Consult GitHub's live models and pricing table and your organization's billing settings before a large team rollout. The amount charged can depend on the product surface and current plan rules, so a price copied into an article can become misleading quickly.

For a fair budget comparison, measure cost per accepted task, not just cost per million tokens or price per request. A model that needs fewer retries may be cheaper overall even if its individual call costs more. The opposite is also possible for straightforward edits. Record the source of every billing figure and the date you checked it.

A five-task comparison for your own codebase

Pick five tasks that look like real work: one small bug fix, one multi-file change, one test repair, one unfamiliar-code explanation and one terminal workflow. Write a short acceptance criterion for each before you run either model. For a bug fix, that might be a failing test that becomes green without altering unrelated behavior. Use the same repository snapshot and prompt for each model.

1. Keep the setup constant. Give both models the same files, permissions and test command.
2. Cap each attempt. Decide in advance how long you will wait and how many follow-up prompts are allowed.
3. Track the outcome. Mark accepted, accepted after edits, or rejected. Count incorrect assumptions and unintended changes.
4. Measure the total. Include human review time, run duration and billed usage, rather than counting generated code lines.
5. Retest a failure. A single unlucky run should not determine a team policy; repeat ambiguous cases with a fresh session.

An example: if both models produce a patch, but one needs twenty minutes of cleanup to pass the repository's tests, the faster initial response may have been the slower task. The same principle applies when an agent chooses a poor terminal command or updates files outside the requested area.

When might Sol be a good fit?

GitHub highlights multistep coding and terminal workflows. That makes a contained repository issue, migration or command-driven diagnosis a reasonable place to test. For a very small completion, your existing default may be good enough. For sensitive repositories, organization policy, allowed tools and code review matter more than a model's name. Never equate a model being available with a patch being ready to merge.

OpenAI's model announcement and API changelog describe a separate API release of GPT-6.1 Sol. Availability and pricing inside GitHub Copilot are governed by GitHub's own product documentation. Do not apply an API price directly to a Copilot invoice.

Frequently asked questions

Why is GPT-6.1 Sol missing from my model picker?

GitHub says the rollout is gradual. Also verify your plan, application version and any administrator model policy.

Is it available on Copilot Pro?

GitHub's September 29 release note lists Pro+, Max, Business and Enterprise. It does not list the standard Pro plan.

Is GPT-6.1 Sol automatically the best coding model?

No single model wins every task. Compare accepted outcomes, review time and cost on your own repository.

Sources and reporting notes

Checked October 3, 2026: GitHub's September 29 release note; GitHub models and pricing; OpenAI model announcement. The five-task evaluation is an editorial method, not a GitHub benchmark.