OpenAI has officially introduced GPT-6 Astra, a new frontier model built for complex reasoning, coding, research, computer use and professional workflows.
Instead of focusing only on answering questions, Astra is designed to complete longer, multi-step tasks across software, websites, documents, spreadsheets and connected tools. OpenAI describes it as its most capable model for difficult end-to-end work.
GPT-6 Astra is becoming available across ChatGPT, ChatGPT Work, Codex and the OpenAI API. However, availability differs by subscription plan and product, and its rollout is gradual.
Here is what GPT-6 Astra offers, who can access it, how much its API costs, and where it may provide a meaningful improvement over previous models.
What Is GPT-6 Astra?
GPT-6 Astra is OpenAI’s flagship model for handling complicated tasks that require reasoning, planning and interaction with digital tools.
In ChatGPT, eligible users may see it presented as GPT-6 Pro, which is powered by the underlying GPT-6 Astra model. In Codex and the API, developers may encounter the more direct Astra model name.
It is designed for work that goes beyond generating a quick paragraph or answering a general question. Astra can analyse requirements, use tools, interact with websites or software, generate files, test its work and adjust when a task changes.
In simple terms: GPT-6 Astra is designed to act less like a basic chatbot and more like an intelligent digital collaborator capable of completing multi-step professional work.
GPT-6 Astra Key Specifications
| Model name | GPT-6 Astra |
|---|---|
| ChatGPT name | GPT-6 Pro on eligible plans |
| Best suited for | Complex reasoning, coding, computer use, research and document creation |
| Context window | 1,050,000 tokens through the OpenAI API |
| Maximum output | 128,000 tokens |
| Knowledge cutoff | April 30, 2026 |
| Image input | Supported |
| Audio input | Not supported by the listed Astra API model |
| Fine-tuning | Not supported |
| API model ID | gpt-6-astra |
These technical details apply to the listed OpenAI API model. Features and limits inside ChatGPT may differ depending on the user’s plan, selected mode and available tools.
What Makes GPT-6 Astra Different?
The most important change is not simply that Astra can produce smarter answers. Its bigger advantage is the ability to remain focused while completing a difficult workflow involving several tools and stages.
For example, a conventional chatbot may explain how to build an application. Astra is intended to go further by reviewing an existing project, changing the code, running tests, checking the interface and fixing problems discovered during verification.
Its main areas of improvement include:
- Complex reasoning and problem-solving
- Software engineering and codebase understanding
- Computer and browser interaction
- Scientific and technical research
- Professional document creation
- Understanding instructions and task boundaries
- Maintaining direction during long tasks
- Making better visual and layout decisions
More Advanced Computer Use
GPT-6 Astra is designed to interact with software and websites when the necessary tools and permissions are available.
OpenAI says Astra can help with activities such as:
- Completing online forms
- Updating customer information in CRM software
- Organising calendar entries
- Conducting online research
- Creating reports in document editors
- Analysing scientific data
- Generating plots and visualisations
- Building and testing websites
- Installing and troubleshooting software
This does not mean Astra automatically receives unrestricted access to a user’s computer. Access depends on the product, available tools, account permissions and confirmation requirements.
OpenAI reports that in its OSWorld 2.0 latency simulations, Astra completed computer-use tasks in approximately 47% less time than GPT-5.6 Sol while also achieving a higher score. Real-world speed will still depend on the task, selected reasoning effort and tool response times.
GPT-6 Astra for Coding and App Development
For developers, coding is one of Astra’s most important improvements.
The model is designed to understand an existing repository, follow project conventions, implement features, find bugs and verify that its changes work. It can also communicate progress while completing longer tasks.
Developers may use Astra for:
- Building mobile and web applications
- Understanding large codebases
- Fixing complex bugs
- Writing and updating automated tests
- Refactoring existing features
- Working across frontend and backend code
- Reviewing pull requests
- Finding performance problems
- Running browser-based quality checks
- Creating technical documentation
For a Flutter developer, this could include inspecting multiple Dart files, implementing a purchase flow, checking Android and iOS configuration, running tests and fixing issues without losing track of the original feature requirements.
Developer reminder: Even a highly capable coding model can make mistakes. Review security-sensitive code, payments, authentication, database migrations and production configuration before deployment.
Better Context Handling in Long Coding Sessions
Long-running development tasks often generate a large amount of context. The conversation may include source code, terminal results, test failures and several attempted fixes.
OpenAI says Astra introduces improvements to how Codex preserves and retrieves information when a working session becomes very long.
This can help the model remember:
- Earlier project requirements
- Why a previous solution failed
- Test results from an earlier step
- Existing architectural decisions
- User instructions added during the task
Better context retrieval may reduce the need to repeatedly explain the same project or correct a model that has forgotten an earlier constraint.
Documents, Presentations and Spreadsheets
GPT-6 Astra is also trained for professional knowledge work.
According to OpenAI, it is better at following existing templates and producing structured documents, spreadsheets, analyses and presentations. Rather than replacing a company’s preferred format with a generic design, it can attempt to follow the supplied style and layout.
Potential uses include:
- Creating business reports
- Preparing structured presentations
- Analysing spreadsheet data
- Writing proposals and project plans
- Formatting professional documents
- Updating an existing template
- Creating charts from business data
- Summarising research into a usable document
Astra also brings improved visual judgement to websites, applications, games and other generated assets. That does not mean every design will automatically be perfect, but the model is intended to make more appropriate decisions about layout and presentation.
Research, Mathematics and Science
OpenAI is positioning Astra as a major step forward for scientific and mathematical work.
The company reports strong results across advanced evaluations covering mathematics, science, biology, medicine and technical research. OpenAI also says Astra contributed to work involving long-standing mathematical problems related to prime-number gaps.
Researchers could use the model to inspect datasets, write analysis code, generate plots, review evidence and operate specialised software when suitable tools are available.
These capabilities should still be used with expert oversight. A high benchmark score does not guarantee that every medical, mathematical or scientific answer will be correct.
Cybersecurity Capabilities and Safeguards
Astra has significantly stronger cybersecurity capabilities than earlier OpenAI models. OpenAI reports that it meets the Critical cybersecurity capability threshold under its Preparedness Framework.
That capability can help security teams review code, identify weaknesses and develop patches. It also introduces obvious misuse risks.
OpenAI says the released version includes stronger safeguards and may refuse advanced requests involving harmful exploit creation. Access to some defensive cybersecurity workflows may also depend on additional programmes, monitoring and account eligibility.
Developers should therefore not assume that every security-related request will be completed, even when it is framed as research.
Is GPT-6 Astra Available in ChatGPT?
Yes, but availability depends on the user’s plan and the ChatGPT product being used.
OpenAI’s current help documentation says GPT-6 Pro, powered by GPT-6 Astra, is rolling out in ChatGPT for eligible Pro, Business and Enterprise plans.
Plus subscribers receive Astra in ChatGPT Work and Codex as it becomes available. Availability can differ between regular Chat, Work and Codex during the rollout.
| Plan or product | Expected Astra access |
|---|---|
| ChatGPT Plus | Included in ChatGPT Work and Codex during rollout; regular Chat availability may differ |
| ChatGPT Pro | GPT-6 Pro access during rollout, subject to usage limits |
| ChatGPT Business | GPT-6 Pro access with plan-specific limits |
| ChatGPT Enterprise | Access depends on rollout and workspace administrator permissions |
| OpenAI API | Available using the gpt-6-astra model, subject to API access and billing |
Not seeing Astra? The rollout is gradual. Check your plan, workspace permissions and the product you are using. Desktop users should also update the ChatGPT application to the latest version.
GPT-6 Astra in Codex
Codex users can select Astra for difficult software-engineering and computer-use tasks when it is available for their account.
For Codex CLI, OpenAI states that Astra requires version 0.153.0 or newer.
The model can be selected with:
If the model is not available after updating Codex, check whether access has reached your plan or whether a workspace administrator needs to enable it.
GPT-6 Astra API Pricing
OpenAI’s published API pricing for GPT-6 Astra is based on tokens.
| Usage type | Price per 1 million tokens |
|---|---|
| Input tokens | $10 |
| Cached input | $1 |
| Cache writes | $12.50 |
| Output tokens | $50 |
Prompts containing more than 272,000 input tokens have higher pricing. OpenAI states that these requests are charged at twice the input and cache rates and 1.5 times the output rate for the complete request.
Batch and Flex processing may cost less than Standard processing, while Fast mode costs more. Tool usage, such as web search or computer use, may introduce separate charges.
Important: A ChatGPT subscription and OpenAI API billing are separate. Paying for ChatGPT Plus or Pro does not automatically provide unlimited free API usage.
GPT-6 Astra Context Window
GPT-6 Astra supports a context window of up to 1,050,000 tokens through the API.
A large context window allows developers to provide substantial amounts of information in one request, such as:
- Large software repositories
- Long legal or business documents
- Research papers and technical reports
- Extended conversation histories
- Large datasets represented as text
- Several related files and project requirements
The model supports a maximum output of 128,000 tokens. Actual usable limits may vary depending on the endpoint, account limits and product configuration.
A larger context window is useful, but passing unnecessary data can increase both cost and processing time. Developers should still provide only information relevant to the task.
Supported Astra API Tools and Features
OpenAI lists support for several capabilities through the Responses API, including:
- Web search
- File search
- Image generation
- Code Interpreter
- Hosted shell
- Apply Patch
- Skills
- Computer use
- Model Context Protocol connections
- Tool search
- Function calling
- Structured outputs
- Streaming
The model supports text and image input but does not directly support audio or video input in its listed API specification. Fine-tuning is also not currently supported.
GPT-6 Astra vs GPT-5.6 Sol
| Area | GPT-6 Astra | GPT-5.6 Sol |
|---|---|---|
| Positioning | Flagship model for the hardest end-to-end tasks | Capable model for coding, research and professional work |
| Reasoning | Designed for more difficult multi-step problems | Suitable for many complex everyday workflows |
| Computer use | Stronger performance and judgement | Capable, but behind Astra in OpenAI’s evaluations |
| Coding | Best suited to demanding repository-level work | Strong option when Astra is unnecessary or unavailable |
| Cost | Higher per-token API price | Usually more economical for less demanding tasks |
| Best choice | Complex, high-value work where quality matters most | Regular development and knowledge work requiring a balance of capability and cost |
Astra is not automatically the best choice for every request. A faster and less expensive model may be more practical for short summaries, basic rewriting, simple customer support or high-volume automation.
Who Should Use GPT-6 Astra?
Astra may be especially useful for:
- Software developers working on complex applications
- Businesses automating multi-step workflows
- Researchers analysing technical information
- Security professionals reviewing and protecting software
- Designers building websites and interactive products
- Analysts creating documents and spreadsheets
- Teams working with large repositories or file collections
For normal questions or lightweight writing, its advanced capabilities may be unnecessary. The model is most valuable when a task involves several connected steps, specialised tools or high-quality professional output.
Limitations You Should Know
Despite its improvements, GPT-6 Astra is not perfect.
- It can still generate incorrect information.
- Its ChatGPT usage is limited by plan.
- API usage can become expensive for large prompts.
- Not every user receives access at the same time.
- Some cybersecurity requests are restricted.
- Fine-tuning is not currently supported.
- Tool-based actions depend on permissions and available integrations.
- Critical work still requires human review.
Astra should be treated as a powerful collaborator—not an unquestionable authority.
Frequently Asked Questions
What is ChatGPT Astra?
ChatGPT Astra commonly refers to GPT-6 Pro powered by OpenAI’s GPT-6 Astra model. It is designed for complex reasoning, coding, computer use, research and professional workflows.
Is GPT-6 Astra officially launched?
Yes. OpenAI officially announced GPT-6 Astra in September 2026. Access is rolling out gradually across eligible products and plans.
Can ChatGPT Plus users access Astra?
OpenAI says Plus plans include Astra in ChatGPT Work and Codex as it rolls out. Availability in standard Chat may differ from Work and Codex.
Is GPT-6 Astra free?
No. ChatGPT access depends on an eligible paid plan, while API access is billed separately according to token and tool usage.
What is the GPT-6 Astra context window?
The official API model page lists a context window of 1,050,000 tokens and a maximum output of 128,000 tokens.
How much does the GPT-6 Astra API cost?
The published price is $10 per million input tokens, $1 per million cached input tokens and $50 per million output tokens. Additional rules apply to very large prompts and tool use.
Is Astra better than GPT-5.6 Sol?
OpenAI positions Astra as its more capable model for difficult end-to-end work. GPT-5.6 Sol may remain a better value for less demanding or higher-volume tasks.
Can GPT-6 Astra create websites and applications?
Yes. When used with appropriate tools, Astra can build websites and applications, modify code, run tests and perform frontend quality checks.
Final Thoughts
GPT-6 Astra represents an important change in how OpenAI expects people to use ChatGPT and Codex.
The focus is moving beyond receiving answers. Astra is designed to work through entire tasks: understanding requirements, using software, editing files, verifying results and adapting when instructions change.
Its strongest applications will likely be software development, research, professional document creation and computer-based automation. However, its higher API pricing and limited ChatGPT availability mean it should be selected for work that genuinely benefits from its additional capability.
For developers, creators and businesses, the most interesting question is no longer only, “What can the model tell me?” It is becoming, “What complete piece of work can the model help me finish?”
