GPT-6 Astra launch resources

What’s new with GPT-6 Astra

GPT-6 Astra is OpenAI’s most intelligent model, combining deeper reasoning and judgment with computer use to take on demanding professional work.

• Deeper judgment. Work through conflicting sources and tradeoffs (e.g., to turn quarterly results into a clear business review).
• Computer use. Gather context and carry work across apps and files with the tools and permissions available in your workspace.
• Professional quality. Create clear writing, polished presentations, and analyses that fit your brief, audience, and templates.

GPT-6 Astra will be made available in a phased rollout beginning September 3, 2026. See the launch announcement for details. Confirm your organization’s access with your workspace admin and applicable usage terms with your OpenAI account team.

AI Platform Admins

Enable access and manage controls, usage, and support.

1. Confirm availability and set  access. Astra is off by default in ChatGPT Enterprise. Once it’s available to your workspace, use its dedicated model permission to choose which users and groups can use it. Verify access in the ChatGPT, Work, or Codex experiences those teams use.
2. Check existing controls with Astra. For workflows that use connected apps or computer use, test Astra with the team’s existing permissions. Confirm it can retrieve approved files, restricted actions remain blocked, and required approval prompts appear. Work with workflow owners to resolve access or configuration issues.
3. Review the impact of usage. Confirm the usage terms that apply to Astra under your plan and review existing user/group limits, workspace overage settings, and alerts. After access is enabled, investigate changes in usage and cost, reached limits, and Astra-related help requests.

Executive Sponsors

Lead by example and shape how your organization uses Astra.

1. Try Astra on a leadership task. Once Astra is available, use it for a familiar business review or decision. Compare the results with how you do that work today, then share what helped, what needed correction, and how you reviewed the result.
2. Give teams time to evaluate Astra. Ask what Astra should improve—such as the rework required for a recurring analysis—and give teams time to compare it with their current approach. Recognize useful findings, including when the current model remains the better choice.
3. Revisit priorities using results. Use results from teams’ Astra comparisons to decide whether a postponed priority or different approach to existing work is now practical. Commit people, time, and investment where the evidence supports a change.

Transformation Leaders

Lead Astra deployment and adoption across the organization.

1. Choose where to evaluate Astra. Identify where Astra could improve work that involves conflicting sources, complex tradeoffs, or demanding deliverables. Align Admins, workflow owners, and stakeholders on which opportunities to assess, who owns them, and what results would justify adopting Astra.
2. Prepare teams for the model change. Confirm availability with Admins before scheduling team access, then explain when to choose Astra. Share tested examples and any changes to instructions or review steps. Name who approves workflow changes and handles questions.
3. Measure the value of adopting Astra. Bring together results from teams using Astra: quality, review and correction time, completion time, usage or cost, recurring use, and business outcomes. Decide with workflow owners which model changes to keep, revise, or reverse.

Agent Activators

Identify which workflows should adopt Astra and support the change.

1. Choose a workflow Astra could improve. Use recurring work where Astra’s judgment or output quality could address a specific problem, such as conflicting sources or repeated corrections. Decide what would justify changing models (e.g., fewer factual errors or less time editing the result).
2. Compare Astra with the current model. Keep a copy of the current version. Check that you can select Astra for the task or agent. Then, test the same examples, prompts, files, tools, permissions, and review standards with Astra and the current model. Compare output quality, actions taken, corrections, completion time, and usage or cost.
3. Recommend whether to switch. Share the comparison and recommend keeping the current model, testing further, or adopting Astra for the work that benefits. Share tested instructions, review steps, and how to return to the previous version.