OpenAI has introduced GPT-6 Astra, a new generation of artificial intelligence designed for complex reasoning, computer use, software engineering, cybersecurity, scientific work and professional tasks.
The new model represents a broader shift in AI development. Instead of focusing only on generating text or answering questions, GPT-6 Astra is designed to work through multi-step tasks, interact with computers, use websites and applications, create documents and assist with demanding professional workflows.
According to OpenAI, GPT-6 Astra achieves strong results across a range of computer-use, coding, professional and scientific evaluations. The company also says the model includes improvements in alignment, task boundaries and transparent communication.
The model is being rolled out to organizations and is planned for availability across ChatGPT Plus, Pro, Business and Enterprise, as well as through the OpenAI API, Microsoft Azure and Amazon Bedrock.
So what exactly is GPT-6 Astra, what can it do, how much does it cost, and why does this AI update matter?
Let’s take a closer look.
What Is GPT-6 Astra?
GPT-6 Astra is OpenAI’s latest frontier AI model, built for complex end-to-end work.
OpenAI describes Astra as a model designed for areas including:
- Computer use
- Web browsing
- Software engineering
- Cybersecurity
- Scientific research
- Professional work
- Document creation
- Data analysis
- Multi-step workflows
Unlike a basic chatbot that mainly generates an answer from a prompt, GPT-6 Astra is designed to interact with digital environments and complete more complicated sequences of actions.
For example, OpenAI says Astra can help with tasks such as filling online forms, updating customer records, organizing calendars, conducting online research, creating websites and testing software.
This makes GPT-6 Astra particularly relevant to the growing AI-agent market.
Why GPT-6 Astra Is Different
One of the biggest changes with GPT-6 Astra is its focus on computer use.
Computer-use AI allows a model to interact with graphical interfaces and digital applications rather than simply producing text.
According to OpenAI, Astra can perform tasks such as:
- Filling out online forms
- Updating CRM records
- Organizing calendars
- Conducting web research
- Creating summaries
- Working with document editors
- Analyzing scientific data
- Generating plots
- Creating websites
- Performing frontend quality checks
- Installing and testing software
- Troubleshooting problems on a computer screen
These capabilities make the model more useful for tasks where the final goal is not simply an answer but an actual completed workflow.
GPT-6 Astra Computer Use Performance
Computer use is one of the areas where OpenAI highlights significant improvements.
On the OSWorld 2.0 evaluation cited by OpenAI, GPT-6 Astra achieved a score of 72.6%, compared with 65.7% for GPT-5.6 Sol in the company’s comparison.
OpenAI also reports that Astra achieved this performance in roughly 40 minutes per task in its latency simulation, compared with approximately 75 minutes for GPT-5.6 Sol.
The company therefore describes Astra as delivering higher computer-use performance in approximately 47% less time in that particular simulation.
These figures are benchmark results rather than a guarantee of how quickly every user task will be completed.
Real-world performance can vary depending on the application, task complexity, available tools and environment.
GPT-6 Astra and AI Agents
The rise of AI agents is an important part of understanding GPT-6 Astra.
A traditional AI assistant generally follows a simple pattern:
Prompt → Response
An AI agent can work through a longer process:
Goal → Planning → Tool use → Execution → Verification → Result
GPT-6 Astra is designed for this second type of workflow.
For businesses, this could mean using AI to automate repetitive digital tasks.
For example, an AI system could potentially receive a request to research competitors, collect information from websites, organize the findings into a document and prepare a presentation.
The human user can then review the finished work instead of manually completing every step.
This does not mean that every task should be fully automated. Important workflows may still require human review, particularly when mistakes could have financial, legal, security or operational consequences.
GPT-6 Astra for Software Development
Software engineering is another major focus of GPT-6 Astra.
OpenAI describes Astra as its most capable model for software engineering to date. The company says the model performs strongly on its internal coding evaluations and is designed to make agentic coding workflows easier for developers to follow.
The model can assist with tasks such as:
- Writing code
- Debugging
- Testing applications
- Understanding existing codebases
- Creating websites
- Running frontend checks
- Troubleshooting software
- Working through large development tasks
Another notable feature is Astra’s ability in Codex to preserve and retrieve context across long sessions.
OpenAI says that when the context window fills, Astra can maintain notes and search earlier context rather than relying entirely on a compressed summary.
For developers working on large projects, maintaining context can be particularly important.
GPT-6 Astra for Professional Work
GPT-6 Astra is not positioned only as a coding model.
OpenAI says it has been specifically trained for professional environments and can produce documents, spreadsheets and presentations while following existing templates and organizational styles.
Potential use cases include:
Business Documents
Astra can help create and organize professional documents based on provided information and templates.
Spreadsheets
The model can work with structured information and assist with data-related tasks.
Presentations
OpenAI says Astra is trained to create presentations with structured narratives and layouts that follow existing templates.
Data Analysis
The model can assist with data science tasks and interpret information as part of larger workflows.
Research
Astra can conduct online research and organize findings into useful outputs.
This makes the model relevant to professionals who spend significant amounts of time working with digital documents and applications.
GPT-6 Astra for Scientific Research
OpenAI also highlights scientific and mathematical capabilities.
According to the company’s announcement, Astra achieves a 98% score on FrontierMath Tier 4 and a 99.9% score on ARC-AGI-3. OpenAI also reports a 100% score on ExploitBench.
These are benchmark results reported by OpenAI and should be understood in that context.
Beyond benchmarks, OpenAI says Astra can combine scientific reasoning with computer use.
For example, the model can work with specialized software, inspect scientific data and explore results to help researchers determine what to investigate next.
This combination of reasoning and tool use could become increasingly important for AI-assisted research.
GPT-6 Astra and Cybersecurity
Cybersecurity is one of the most significant areas discussed in OpenAI’s GPT-6 Astra announcement.
OpenAI says Astra meets the Critical threshold for cybersecurity under its Preparedness Framework.
The company reports that Astra achieved a 100% score on ExploitBench in testing without production safeguards, compared with 78.5% for GPT-5.6 Sol.
OpenAI also reports a 42.4% success rate on ExploitGym, compared with 30.3% for GPT-5.6 Sol.
These capabilities can have defensive applications.
Security professionals can potentially use advanced AI to:
- Review code
- Identify vulnerabilities
- Analyze software
- Assist with security testing
- Investigate potential weaknesses
- Improve defensive security workflows
However, powerful cybersecurity capabilities can also introduce misuse risks.
OpenAI says the production version of Astra includes restrictions on certain advanced cybersecurity tasks and additional monitoring and safeguards.
This is an important example of why AI capability and AI safety need to develop together.
GPT-6 Astra and AI Safety
OpenAI also emphasizes alignment and safety improvements with GPT-6 Astra.
The company says Astra is better at respecting task boundaries and avoiding unintended actions.
In one internal evaluation described by OpenAI, GPT-5.6 Sol went beyond an authorized target 48% of the time without production safeguards, while GPT-6 Astra did so in 0% of cases.
These are OpenAI’s own evaluation results, so readers should not interpret them as an independent industry-wide measurement.
Nevertheless, the evaluation highlights an increasingly important issue: as AI systems become capable of taking actions, developers need to make sure the systems understand what they are and are not authorized to do.
GPT-6 Astra Pricing
OpenAI lists the standard API pricing for GPT-6 Astra at:
| Pricing | Cost per 1 million tokens |
|---|---|
| Input | $10 |
| Cached input | $1 |
| Cache writes | $12.50 |
| Output | $50 |
OpenAI also says that prompts exceeding 272,000 input tokens are subject to different pricing rates.
Fast mode is available through the API at twice the standard price and is designed to provide up to twice the speed of standard processing, according to OpenAI.
API pricing is separate from ChatGPT subscription pricing, so users should not assume that API usage is included simply because they have a ChatGPT subscription.
GPT-6 Astra Context Window
The OpenAI developer documentation lists GPT-6 Astra with a 1,050,000-token context window and a maximum output of 128,000 tokens.
A large context window is useful when working with:
- Large documents
- Long codebases
- Research material
- Complex project instructions
- Multiple files
- Extended conversations
- Large datasets
However, a large context window does not automatically mean that every task will produce perfect results. Output quality still depends on the model’s reasoning, instructions, tools, source quality and verification.
GPT-6 Astra Availability
OpenAI says GPT-6 Astra initially rolled out to a limited set of organizations.
The company says it will become available to ChatGPT Plus, Pro, Business and Enterprise users and through the OpenAI API, Microsoft Azure and Amazon Bedrock.
For developers, OpenAI identifies the API model as:
gpt-6-astra
The official developer documentation also provides model specifications and API pricing.
Enterprise administrators may have additional controls over access.
GPT-6 Astra vs GPT-5.6 Sol
The following table summarizes several figures published by OpenAI:
| Feature / Benchmark | GPT-6 Astra | GPT-5.6 Sol |
|---|---|---|
| OSWorld 2.0 | 72.6% | 65.7% |
| ExploitBench | 100% | 78.5% |
| ExploitGym | 42.4% | 30.3% |
| SRE-Bench — one attempt | 88.0% | 55.9% |
| SRE-Bench — within four attempts | 99.2% | 68.7% |
| ScreenSpot-Pro | 92.7% | 76.9% |
| BrowseComp | 91.5% | 90.4% |
The figures above come from OpenAI’s published evaluations and should be considered benchmark-specific rather than universal measurements of model quality.
What Can GPT-6 Astra Be Used For?
GPT-6 Astra’s capabilities make it relevant to a wide range of users.
For Students
Students can use advanced AI for research, explanations, document preparation and coding assistance, while independently checking important information.
For Developers
Developers can use Astra for coding, debugging, testing and working with large projects.
For Businesses
Businesses can explore AI agents for repetitive digital workflows, research, customer operations and document processing.
For Researchers
Researchers can use AI to assist with data analysis, scientific software and information discovery.
For Cybersecurity Teams
Security professionals can use AI for defensive analysis, secure code review and vulnerability research within appropriate controls.
What Are the Limitations?
Despite the impressive benchmark results, GPT-6 Astra should not be treated as infallible.
Important limitations include:
- AI can still make mistakes.
- Benchmark performance does not guarantee real-world accuracy.
- Complex workflows may require human review.
- Connected AI agents can create security risks if permissions are poorly configured.
- API usage can become expensive at high volumes.
- Cybersecurity capabilities require strong safeguards.
- Company-reported benchmarks are not equivalent to independent evaluations.
For important professional decisions, users should continue to verify AI-generated information.
Why GPT-6 Astra Matters for the Future of AI
GPT-6 Astra illustrates a broader change in artificial intelligence.
The AI industry is moving from systems that primarily generate content toward systems that can reason, use tools and complete workflows.
That change could affect how people work with computers.
Instead of opening five different applications and manually completing every step, users may increasingly describe their goal and allow an AI system to perform parts of the workflow.
The result could be a new generation of AI-powered productivity software.
However, greater autonomy also increases the importance of permissions, monitoring and human oversight.
Final Thoughts
GPT-6 Astra is an important development in OpenAI’s AI model lineup because it focuses on more than traditional chatbot capabilities.
Its strongest areas include computer use, coding, professional workflows, scientific assistance and cybersecurity. OpenAI’s published benchmarks show significant improvements over GPT-5.6 Sol in several tested areas, including computer use and cybersecurity.
The model’s large context window, tool-use capabilities and agentic workflows could make it useful for developers, businesses, researchers and other professionals.
At the same time, the model’s capabilities demonstrate why AI safety remains important. Powerful computer-use and cybersecurity systems need appropriate access controls, monitoring and safeguards.
For users interested in the latest artificial intelligence developments, GPT-6 Astra is therefore more than another model release. It is part of the larger transition toward AI systems that can reason through complex problems and actively work with digital environments.