GitHub announced general availability of GPT-6 Astra in GitHub Copilot on September 4, 2026. The rollout adds another option for teams evaluating AI-assisted debugging and larger automation changes.
What GitHub announced
According to GitHub’s official announcement, Astra targets extended autonomous coding tasks. GitHub reports that its internal tests showed stronger long-task performance with fewer steps than earlier OpenAI models, including ongoing planning and result verification. These are vendor-reported findings, not a published QA benchmark from QATechTools.
Access covers Copilot Pro+, Max, Business, and Enterprise. Supported surfaces include VS Code, Visual Studio, Copilot CLI, the coding agent, the Copilot app, github.com, mobile, and several other IDEs. Availability is gradual, so the model may not appear immediately.
Business and Enterprise administrators control access through model policy. GitHub says new models are enabled by default unless administrators change that setting or disable the specific model. Astra uses provider list pricing under usage-based billing.
Why this matters for QA engineers
Our QA takeaway: longer coding runs deserve evaluation at the workflow level. A useful trial is a reproducible defect that requires diagnosis, a regression test, a repair, and a clean rerun. Completion claims alone do not show that the original failure is fixed.
Before expanding adoption, compare Astra with your current model using the same repository revision and defect set. Record whether the new test fails before the patch and passes afterward, whether unrelated tests regress, and how much reviewer correction is needed. Track elapsed time and usage cost alongside correctness.
Keep CI results and human review as independent acceptance evidence. Start with a small set of known failures; expand only after the generated changes meet your existing release criteria.
News date: September 4, 2026. Verified September 6, 2026. QA evaluation recommendations above are editorial analysis.