GPT-6 Astra Is Now in GitHub Copilot for Pro+, Business, and Enterprise

OpenAI's GPT-6 Astra, the company's latest general-purpose model built for autonomous, long-horizon coding tasks, became generally available inside GitHub Copilot on September 4, 2026. Access is rolling out gradually to Copilot Pro+, Max, Business, and Enterprise subscribers across VS Code, Visual Studio, JetBrains IDEs, Xcode, Eclipse, the Copilot CLI, GitHub Mobile, and github.com. Billing runs at OpenAI's provider list pricing under Copilot's usage-based billing system.

What happened

Detail	Fact
Model	GPT-6 Astra (OpenAI)
Available from	September 4, 2026
Eligible plans	Copilot Pro+, Max, Business, Enterprise
Billing	Provider list pricing, usage-based
Rollout	Gradual. Not all users will see it immediately.

GPT-6 Astra is OpenAI’s latest general-purpose model, designed specifically for long-horizon and agentic coding work. “Agentic” here means the model can carry out multi-step tasks on its own: it plans, validates, and batches diagnosis with verification steps rather than waiting for the user to guide each move. According to GitHub’s internal testing, it independently confirms its own results before declaring a task complete, and it handled long-horizon coding tasks in fewer steps than prior OpenAI models.

You can select GPT-6 Astra from the model picker in all major Copilot surfaces:

• Visual Studio Code and Visual Studio
• JetBrains IDEs, Xcode, and Eclipse
• Copilot CLI and the GitHub Copilot coding agent
• GitHub.com and GitHub Mobile (iOS and Android)
• The GitHub Copilot app

How admins control access

For Copilot Business and Enterprise plans, administrators manage GPT-6 Astra through the model policy inside Copilot settings. Under GitHub’s default model enablement behaviour, new models are switched on automatically unless the administrator has disabled the global default or explicitly blocked this specific model. If your organisation restricts AI model access, check that policy before assuming the model is live for your team.

Why it matters

GPT-6 Astra is aimed squarely at the use cases where earlier Copilot models stalled: large refactors, multi-file changes, and tasks that require the model to reason about its own output before finishing. If GitHub’s internal benchmark holds in production, fewer manual back-and-forth review cycles on agentic tasks is a real workflow improvement, not just a spec sheet upgrade.

For teams already paying for Copilot Business or Enterprise, this is an included upgrade path (subject to usage-based billing costs). The provider list pricing means costs scale with use, so heavy users on flat-rate plans should check the AI integration implications before enabling it org-wide without caps in place.

It is also worth noting that Astra was previously flagged by OpenAI as the first model rated at “critical” cyber risk under its own preparedness framework. GitHub’s changelog makes no mention of any access restrictions tied to that classification, but enterprise security teams should be aware of that context.

Our take

The headline feature, self-validation before marking a task done, is genuinely interesting. Most current models will confidently return broken code. A model that checks its own output first could meaningfully reduce the review burden on senior developers. That said, GitHub’s numbers come from internal testing, not an independent benchmark. We would treat the “fewer steps” claim as a promising signal to verify in your own codebase, not a guaranteed throughput gain.

For businesses evaluating whether to enable this model across their org, the usage-based billing is the thing to watch. Long-horizon agentic tasks consume more tokens than a quick autocomplete. If your team starts using GPT-6 Astra for large refactors, costs can compound quickly. Set a spending alert before you roll it out broadly.

If you are still choosing a Copilot plan or thinking about integrating AI tools into your development workflow, now is a reasonable moment to revisit the tier comparison, since Astra is not on the base Copilot plan.

What to do about it

1. Check your Copilot plan. GPT-6 Astra requires Pro+, Max, Business, or Enterprise. Base Copilot users do not have access.
2. Look in the model picker inside VS Code, JetBrains, or github.com. If you do not see Astra yet, the rollout is gradual and may take a few more days.
3. Review your admin policy. Business and Enterprise admins should check whether global default model enablement is on or off in Copilot settings.
4. Set a usage alert. Usage-based billing means agentic, long-context tasks will cost more than typical completions. Cap spending before enabling broadly.
5. Test on a real task. Run it against a meaningful refactor or a multi-file feature and compare the output quality and step count against your current default model.

The practical takeaway: enable it selectively, watch the billing, and let your team’s own results, not GitHub’s internal benchmarks, tell you whether it earns a permanent spot in your workflow.

Source: GitHub Changelog