OpenAI GPT-6 Astra: New AI Model Designed to Think, Act and Complete Complex Tasks

OpenAI’s latest model combines advanced reasoning, computer use, coding, science and professional workflows, aiming to move AI beyond simple question-and-answer interactions.

Chennai: OpenAI has introduced GPT-6 Astra, describing it as its most intelligent and aligned AI model yet. The new model is designed to do more than respond to prompts. It can reason through complex problems, use computer tools and work through multi-step tasks from beginning to end.

The launch marks a further shift towards agentic AI, where models can independently plan and execute a series of actions while adapting when requirements change.

GPT-6 Astra Can Act Like a Computer Operator

One of the key features of GPT-6 Astra is its ability to interact with computers and online services. The model can browse websites, fill out forms, update CRM records, organise calendars, create websites and test software.

Rather than providing instructions for users to follow manually, Astra is designed to carry out multiple steps itself. It can plan a task, use different tools and adjust its approach as circumstances change.

Stronger Reasoning and Problem-Solving

GPT-6 Astra also focuses heavily on advanced reasoning and mathematical problem-solving. OpenAI says the model scored 99.9% on ARC-AGI-3 and 97.6% on FrontierMath Tier 4.

The company says Astra has contributed to work on difficult mathematical problems, including research involving both short and large gaps between prime numbers.

Built for Large Software Projects

Software development is another major area for GPT-6 Astra. The model can understand larger codebases, write and debug programs, run tests and refactor multiple files.

Within Codex, Astra can preserve useful information across context windows. This is intended to help developers maintain continuity when working on long and complicated software projects instead of repeatedly providing the same background information.

Professional and Scientific Applications

Astra is also designed for professional and scientific workflows. It can create documents, spreadsheets and presentations while following existing templates and styles.

For researchers, the model can analyse scientific data, generate plots and work with specialised software. These capabilities can help researchers explore results and identify areas that may require further investigation.

Cybersecurity Brings Both Benefits and Risks

Cybersecurity is among GPT-6 Astra’s most powerful capabilities. OpenAI says the model achieved 100% on its ExploitBench evaluation and reached the company’s “Critical” cybersecurity capability threshold.

Because of these capabilities, OpenAI has added stronger safeguards. The launch version refuses certain advananced cybersecurity tasks and includes monitoring for potentially unsafe activity.

Better Control Over AI Actions

Another area OpenAI highlights is Astra’s ability to respect user-defined task boundaries. In one internal evaluation, the model recorded 0% unauthorised scope violations, compared with 48% for GPT-5.6 Sol.

This is important as AI systems become increasingly capable of performing actions on behalf of users. Keeping those actions within the requested scope is a key part of using agentic AI safely.

GPT-6 Astra Availability

OpenAI says GPT-6 Astra is rolling out through ChatGPT, API, Azure and AWS Bedrock.

With its combination of reasoning, computer interaction, coding and professional capabilities, Astra represents OpenAI’s push toward AI systems that can not only understand complex instructions but also take action and complete the work.