AI Catchup

OpenAI Launches GPT-6 Astra for Computer Use and Complex Work

By 8 min read

OpenAI launched GPT-6 Astra on September 3, 2026, for computer use and complex work. Its October 16 update to GPT-6 Pro in ChatGPT adds faster responses, improved plugin and tool use, and parallel agents, rolling out to eligible Pro users.

Update: On October 16, GPT-6 Pro in ChatGPT gained faster responses, improved plugin and tool use, and parallel agents. OpenAI says the update is rolling out to eligible Pro users (release notes).

OpenAI launched GPT-6 Astra on September 3, 2026, as a new model aimed at computer use, browsing, software engineering, cybersecurity, science, and professional work. The rollout started with a limited set of organizations (OpenAI on X; OpenAI) and has since reached every surface OpenAI named: the OpenAI API (OpenAI API changelog), ChatGPT Work and Codex (Codex pricing), Microsoft Foundry on Azure (Microsoft Azure Blog), and Amazon Bedrock (AWS model card).

The launch is a meaningful workflow update rather than only a model-name change. OpenAI describes Astra as a model that can operate across multistep computer tasks, create and edit professional documents, and combine reasoning with tool use. (OpenAI; OpenAI Help Center)

GPT-6 Pro adds parallel agents in ChatGPT

OpenAI's October 16 release notes say GPT-6 Pro in ChatGPT's chat tab now responds faster, uses plugins and tools more effectively, and can split parts of a task across multiple agents working in parallel. Users can follow each agent's progress and open completed tasks to review the work. The update is rolling out to eligible Pro users; availability in managed workspaces depends on workspace settings and rollout eligibility. (OpenAI Help Center; OpenAI on X)

OpenAI's GPT-6 model guide says GPT-6 Pro is powered by GPT-6 Astra, and the October 16 release-note entry places this feature update in ChatGPT's chat tab. (GPT-6 models in ChatGPT; OpenAI Help Center)

For users with access, the practical change is that a single ChatGPT task can expose parallel agent work and let you inspect completed results in the conversation. The rollout is plan- and workspace-dependent, so check the model picker and workspace settings rather than assuming the feature is enabled for every account. (OpenAI Help Center)

What Astra is designed to do

OpenAI positions Astra as a general-purpose model for work that crosses reasoning and execution. Its launch examples include:

  • Filling out online forms and updating records in a CRM.
  • Researching online and drafting summaries in email or document editors.
  • Analyzing scientific data and generating plots.
  • Creating a website and checking its frontend behavior.
  • Installing and testing software, then troubleshooting what appears on screen.
  • Producing documents, spreadsheets, and presentations that follow a user's templates and instructions.

The official announcement also describes Astra as an advance in computer use and browser use, with the model able to adapt when requirements change during a task. (OpenAI; OpenAI Help Center)

Rollout and access

Astra first rolled out to a limited set of organizations, and at launch OpenAI's release notes said it was not yet generally available (OpenAI Help Center). It is now on every surface the launch announcement named:

  • OpenAI API: OpenAI's changelog lists GPT-6 Astra as released on September 3, 2026, as gpt-6-astra on the Responses and Chat Completions APIs (OpenAI API changelog; model page).
  • ChatGPT Work and Codex: OpenAI's Codex pricing page estimates 5 to 45 local GPT-6 Astra messages per five hours on Plus and Standard Business, and says Pro plans currently have no five-hour limit (Codex pricing).
  • Microsoft Foundry on Azure: Microsoft says GPT-6 Astra "is now generally available for all customers in Microsoft Foundry" (Microsoft Azure Blog).
  • Amazon Bedrock: AWS's model card lists a launch date of September 8, 2026, and an Active lifecycle status (AWS model card).

Astra usage is included within existing ChatGPT subscription allowances, according to OpenAI. Users and businesses will also be able to purchase additional usage credits, while Pro, Business, and Enterprise plans are slated to receive GPT-6 Astra Pro. Enterprise administrators can enable Astra for their workspace; OpenAI says access is off by default at launch. (OpenAI)

For API developers, the model identifier is gpt-6-astra. OpenAI lists Standard API pricing at $10 per million input tokens and $50 per million output tokens, with separate rates for cache reads and writes. Fast mode targets up to twice the speed of Standard processing at twice the Standard price. (OpenAI)

Capability signals, with the right caveat

OpenAI reports that Astra is state-of-the-art on several internal and public evaluations, including a 98% score on FrontierMath Tier 4, 99.9% on ARC-AGI-3, and 100% on ExploitBench. It also reports a 72.6% score on OSWorld 2.0 at roughly 40 minutes per task, compared with 65.7% at roughly 75 minutes for GPT-5.6 Sol in the cited latency simulation. These are OpenAI-reported results, not independent comparative testing. (OpenAI)

The practical takeaway is that Astra is intended for longer workflows where the model must inspect a screen, use tools, and produce a finished artifact instead of returning only text. Teams should still test their own tasks, especially where a workflow can change data, send messages, or make external commitments.

Safety and monitoring

OpenAI describes Astra as its first model to reach the Critical cybersecurity capability level under its Preparedness Framework. The safety overview says that, with the right tools and access, the model can find previously unknown security flaws and develop exploit paths across well-protected systems. (OpenAI safety overview)

That capability comes with additional controls. OpenAI says Astra is externally deployed with misalignment monitoring across tool-using inference, and its release notes describe monitoring that can pause or stop a conversation when the model may not have interpreted instructions correctly. Users may then review the situation before proceeding. (OpenAI safety overview; OpenAI Help Center)

OpenAI also says eligible API customers can use Zero Data Retention, and that Private Safety Processing is being tested to strengthen safety monitoring while preserving customer privacy. The safety overview separately notes that Astra's monitorability is lower than GPT-5.6 Sol in some adversarial evaluations, so the launch should not be read as a claim that monitoring eliminates all risk. (OpenAI; OpenAI safety overview)

What builders should check first

Before putting Astra into a production workflow, builders should verify:

  1. Whether their ChatGPT plan, workspace settings, or API organization give them Astra access.
  2. Which platform has the features they need: ChatGPT, the OpenAI API, Azure, or Bedrock.
  3. Whether the workflow needs standard or Fast API processing, and how cache pricing applies.
  4. How pauses, reviews, and monitoring interact with long-running tool calls.
  5. What data-retention and administrator controls apply to their deployment.

GPT-6 Astra is therefore both a model launch and an access change: it brings a more execution-oriented model to OpenAI's product and developer surfaces, and the safety controls are part of the product behavior.

Corrections and Updates

  • October 9, 2026: This page said GPT-6 Astra was not yet generally available and was planned for the OpenAI API, Azure, and AWS Bedrock. It is now available in the API, ChatGPT Work and Codex, Microsoft Foundry, and Amazon Bedrock.
  • October 17, 2026: Added OpenAI's October 16 update to GPT-6 Pro in ChatGPT, including parallel agents, faster responses, and improved plugin and tool use.

Sources

Read next

Related coverage

Frequently Asked Questions

What is GPT-6 Astra?

GPT-6 Astra is OpenAI's new model for computer use, browsing, software engineering, cybersecurity, science, and professional work. OpenAI says it can handle multistep workflows and produce documents, spreadsheets, and presentations.

Where is GPT-6 Astra available?

GPT-6 Astra is in the OpenAI API as gpt-6-astra, in ChatGPT Work and Codex on paid plans including Plus and Business, in Microsoft Foundry, where Microsoft calls it generally available, and in Amazon Bedrock, which lists a September 8, 2026 launch date.

How much does GPT-6 Astra cost in the OpenAI API?

OpenAI lists Standard API pricing of $10 per million input tokens and $50 per million output tokens. Cache reads and writes have separate rates, and Fast mode costs twice Standard processing while targeting up to twice the speed.

What safety controls does GPT-6 Astra use?

OpenAI says Astra includes additional safety monitoring for possible instruction misunderstandings, and a conversation may be paused or stopped for review. OpenAI also says Astra reaches its Critical cybersecurity capability threshold and is deployed with misalignment monitoring for tool-using inference.

Get the weekly AI Catchup

Tools, practices, and what matters, in your inbox every week.