What happened
GitHub says GPT-6 Astra from OpenAI is now available in GitHub Copilot for Pro+, Max, Business, and Enterprise users. The model can be selected across several supported clients, including Visual Studio Code, Visual Studio, Copilot CLI, GitHub Copilot coding agent, GitHub Copilot app, GitHub Mobile, JetBrains IDEs, Xcode, and Eclipse. Access is being rolled out gradually.
GitHub announced on September 4, 2026, that GPT-6 Astra is available in GitHub Copilot. GitHub describes Astra as an OpenAI general-purpose model designed for long-horizon autonomous coding and agentic tasks.
According to GitHub, its internal testing found that Astra plans and validates during tasks, batches diagnosis with verification, and independently confirms results before declaring a task complete. GitHub says this produced stronger performance on long-horizon coding tasks with fewer steps than prior OpenAI models. No test set, scores, methodology, or independent results are provided in the source.
Astra is available to Copilot Pro+, Max, Business, and Enterprise users through the model picker in Visual Studio Code, Visual Studio, Copilot CLI, GitHub Copilot coding agent, the Copilot app, GitHub.com, GitHub Mobile on iOS and Android, JetBrains IDEs, Xcode, and Eclipse. GitHub says rollout is gradual, so eligible users may not see the model immediately.
The model uses provider list pricing under usage-based billing. The source does not state the actual price, quotas, rate limits, or whether any clients or plans have different access conditions. Business and Enterprise administrators can manage access through Copilot model policies; automatic enablement depends on organization-wide settings.
Why it matters
The availability change puts OpenAI’s newest general-purpose coding model into a widely used developer workflow, including agentic coding tools. GitHub says its internal testing found stronger performance on long-horizon coding tasks and fewer steps than prior OpenAI models, but those results are GitHub’s own claims rather than independent evaluation. The practical effect will depend on rollout, plan eligibility, usage costs, and how reliably the model validates its work in real repositories.
This is a concrete deployment of a frontier general-purpose model into coding products used by individual developers and organizations. Its inclusion in Copilot’s coding agent and command-line and IDE integrations could make long-running, multi-step software tasks more accessible to users already working inside those tools.
GitHub’s description emphasizes process behavior—planning, verification, and confirmation—not only generated code. If those capabilities hold outside GitHub’s internal testing, they could reduce the number of manual interventions required for complex coding tasks. The evidence supplied here does not establish that outcome independently.
The cost and governance implications remain material unknowns. Usage-based billing may make extended agentic tasks more expensive than short coding interactions, but the source provides no prices or usage estimates. Organizations also retain administrative controls that can limit availability.
What to watch next
Watch for the completion of the gradual rollout, published usage-based pricing details, and independent testing of Astra’s coding and agentic performance. Business and Enterprise administrators should also check model policies, because administrators can enable or disable access and global default settings may automatically enable new models.
GitHub’s gradual rollout may affect when users in each plan, region, client, or organization can select Astra. The source gives no completion date or geographic schedule.
Independent benchmarks and real-world developer reports are needed to assess the internal-testing claims about long-horizon coding, verification, task length, and reliability.
GitHub’s model-pricing documentation should clarify provider list prices, quotas, and any plan-specific restrictions. None of those amounts or limits are included in the announcement.
Enterprise and Business administrators should monitor the interaction between global default model settings and explicit model-policy controls.