GPT-6 Astra reaches GitHub Copilot for long-horizon coding

GitHub has made OpenAI’s GPT-6 Astra generally available in GitHub Copilot. The model is aimed at long-horizon coding and agentic tasks, and GitHub says its rollout covers Copilot Pro+, Max, Business, and Enterprise users across the web, supported IDEs, Copilot CLI, the Copilot coding agent, and mobile.

The release is a new distribution milestone for GPT-6 Astra, separate from OpenAI’s September 3 launch. OpenAI described Astra as a model for computer use, software engineering, cybersecurity, science, and professional work; GitHub is now putting that capability inside a developer workflow where model choice, administrator policy, and usage billing matter as much as benchmark claims.

What GPT-6 Astra adds to GitHub Copilot

GitHub positions Astra around how an agent works through a task. Its announcement highlights planning, validation, batched diagnosis, and independent confirmation before the model reports completion. Those are workflow behaviors for multi-step changes, rather than a promise that every generated patch will be correct.

The model can be selected in Visual Studio Code, Visual Studio, Copilot CLI, the GitHub Copilot coding agent, the Copilot app, github.com, GitHub Mobile, JetBrains IDEs, Xcode, and Eclipse. The availability list is broad, but the rollout is gradual, so the model may not appear in every account immediately.

Plans, billing, and administrator controls

Access is listed for Copilot Pro+, Max, Business, and Enterprise. GitHub says GPT-6 Astra uses provider list pricing under usage-based billing. That makes the model’s effective cost dependent on the requests and plan rules attached to the selected Copilot experience; it should not be read as an unlimited inclusion in a fixed subscription.

Business and Enterprise administrators can manage access through Copilot model policies. GitHub’s default-model setting enables new models automatically unless an administrator has disabled the global default or explicitly disables Astra. Teams with cost or data-governance requirements should check those policies before enabling it broadly.

The security trade-off is part of the release

OpenAI’s release places Astra in a higher cybersecurity capability tier and says production safeguards will refuse some advanced offensive requests. OpenAI also describes stronger monitoring and action controls, while warning that safeguards can pause legitimate work. GitHub’s announcement does not independently validate those safety claims; it establishes the Copilot availability and billing surface.

For engineering teams, the practical question is therefore operational: which users can select Astra, which agent surfaces can invoke it, how usage-based charges are tracked, and what review gates remain in place for autonomous changes. Treat the model as an accelerator for reviewable work, not as an approval mechanism.

Source check

GitHub’s September 4 changelog is the primary source for Copilot availability, supported surfaces, plan eligibility, gradual rollout, billing, and administrator controls. OpenAI’s September 3 release independently documents GPT-6 Astra’s broader model launch, capabilities, safeguards, API identifier, and staged availability through ChatGPT, the API, Azure, and Bedrock.

The next concrete checkpoint is account-level rollout: eligible Copilot users should verify whether Astra appears in the model picker and confirm their organization’s model policy and usage reporting before assigning it to long-running coding agents.