GitHub Adds GPT-6.1 Sol to Copilot for Multistep Coding Work

A developer working in an editor, a terminal or Copilot’s coding agent can now choose the same new OpenAI model as GitHub rolls GPT-6.1 Sol into Copilot. GitHub announced the addition on September 29, 2026, and says the model is generally available. The rollout is gradual, so eligible users may not see it in the model picker yet.

The addition is aimed at coding jobs that take several steps and at work directed through a terminal. GitHub says early tests found Sol reliably completed tasks while using noticeably fewer tokens and steps than earlier GPT-6 and GPT-5.6 models. That is GitHub’s assessment, not a published measure of how much time or money a particular coding job would save.

One model choice across the coding workflow

Eligible Copilot Pro+, Max, Business and Enterprise users can select Sol from the model picker. GitHub lists Visual Studio Code, Visual Studio, JetBrains IDEs, Xcode and Eclipse among the supported editors. It also lists Copilot CLI, the Copilot coding agent, the Copilot app, github.com and GitHub Mobile on iOS and Android.

That range matters for GitHub’s intended use: a multistep job may begin in an editor and continue through command-line work or a coding agent. Sol is offered as a selectable option in each listed surface, rather than a model confined to one part of Copilot. GitHub has not given a date when every eligible account will see it.

A team’s default may decide access

For Business and Enterprise accounts, administrators manage access through model policies in Copilot settings. Under GitHub’s default model enablement, new models are enabled automatically unless an administrator has turned off the global default or explicitly disables this model. A user who cannot select Sol may therefore be waiting for the rollout—or may be subject to an organization’s policy.

Usage is billed at provider list pricing under usage-based billing. GitHub directs customers to its Copilot models and pricing documentation for details; the announcement itself does not quote a rate. Fewer tokens could matter to a usage bill, but GitHub’s early-test description supplies no numerical saving to apply to a team’s own work.

The efficiency claim still needs a local test

GitHub’s comparison combines two useful measures, but they answer different questions. Tokens describe how much text the model processes or produces; steps describe how many moves it takes through a task. GitHub says Sol used fewer of both in early testing while completing the work. It gives no test-by-test results or size of the reduction in the announcement, leaving users to see whether that pattern holds for their own coding jobs.