GitHub Copilot adds OpenAI GPT-6.2 Astra and Sol for agentic coding
GitHub Copilot users are getting two new OpenAI models for agentic coding and code review, but the early performance claims remain company-reported.
GitHub has added OpenAI’s GPT-6.2 Astra and GPT-6.2 Sol to GitHub Copilot, giving subscribers two new model options for agentic coding and code review. The company says the models are generally available now, although the rollout is gradual and some users may not see them immediately.1
Astra is presented as the model for agentic coding and terminal work. GitHub says early testing found that it caught significantly more real issues in code review than GPT-6 Astra, while completing agentic coding tasks with noticeably fewer tokens. Sol is aimed at complex agentic coding and deeper code review; GitHub says it caught nearly twice as many real issues as GPT-6.1 Sol while keeping its findings accurate.1
Those figures are GitHub’s own early-testing results rather than an independent benchmark. The company has not published a full test methodology in the announcement, so they are useful context for the launch but not a reason to treat either model as objectively superior in every codebase. OpenAI’s accompanying announcement provides the first-party context for the two models, while GitHub’s post covers how they are being exposed to Copilot users.2
Availability and pricing
GPT-6.2 Astra and GPT-6.2 Sol are available to Copilot Pro+, Max, Business and Enterprise subscribers. They can be selected across Visual Studio Code, Visual Studio, Copilot CLI, the GitHub Copilot coding agent, the Copilot app, github.com, GitHub Mobile, JetBrains IDEs, Xcode and Eclipse. Business and Enterprise administrators can control access through Copilot’s model policy.1
Usage is billed under provider list pricing, rather than being included as an unlimited extra. GitHub says the rollout will happen gradually, so availability can vary by plan, organisation settings and client. Developers who cannot see the models yet may need to wait for the rollout rather than change a setting.
Our opinion
This is a meaningful expansion of Copilot’s model picker, but the most important claim still needs outside scrutiny. GitHub’s description suggests Astra and Sol are tuned for different kinds of agentic work, which is more useful than presenting them as interchangeable upgrades. The practical test will be whether their review findings remain reliable on ordinary repositories, not just on the company’s early test set. Until independent comparisons arrive, developers should treat the new models as options to evaluate against their own code, latency and usage budgets rather than automatic replacements for every existing Copilot workflow.