GPT-6 Astra: OpenAI Introduces Its Most Advanced AI Model Yet

OpenAI has introduced GPT-6 Astra, a new AI model designed to handle complex tasks with greater accuracy, speed, and reliability. The company says Astra is its most intelligent and aligned model so far, with major improvements in computer use, coding, scientific research, cybersecurity, and professional work.

GPT-6 Astra combines advances in pre-training, reinforcement learning, and alignment research. It is built to do more than simply answer questions. The model can interact with computers, browse the web, work with software, analyze data, and complete multi-step tasks with less human involvement.

GPT-6 Astra: OpenAI Introduces Its Most Advanced AI Model Yet

The model has also achieved strong results across several demanding benchmarks. OpenAI says Astra scored 98% on FrontierMath Tier 4 and 99.9% on ARC-AGI-3. It also achieved a perfect 100% score on ExploitBench. These results show the model’s focus on solving difficult problems rather than only improving traditional language-based performance.

Better Computer Use

One of the biggest improvements in GPT-6 Astra is its ability to use computers. The model can fill out online forms, update customer records, organize calendars, conduct online research, and prepare summaries in documents or email.

It can also work with technical software. For example, Astra can analyze scientific data, create charts, build websites, test frontend features, and troubleshoot software problems. OpenAI says these capabilities make the model more useful for tasks that normally require several manual steps.

On the OSWorld 2.0 benchmark, Astra achieved a score of 72.6%, compared with 65.7% for GPT-5.6 Sol. OpenAI also says Astra can complete similar computer-use tasks in significantly less time.

Stronger Professional and Coding Skills

GPT-6 Astra is also aimed at professional users. It can create and edit documents, spreadsheets, presentations, and other work-related files. The model is designed to follow existing templates and understand the style and structure of business documents.

Coding is another major focus. On Terminal-Bench 4.0, Astra achieved 57.9%, compared with 37.3% for GPT-5.6 Sol. The model can work with codebases, run tests, use terminal tools, and handle complicated software engineering tasks.

Astra also introduces improved context handling in Codex. Instead of repeatedly compressing older information into summaries during long sessions, it can preserve notes and search earlier context when needed.

Advances in science and cybersecurity

OpenAI says Astra can assist researchers with scientific workflows. It can use specialized software, inspect datasets, run simulations, and help researchers decide what to investigate next. It achieved 96% on the GPQA Diamond benchmark for graduate-level scientific reasoning.

Cybersecurity is another area where Astra has made major progress. OpenAI reports a 100% score on ExploitBench and stronger performance on several security evaluations. Because these capabilities can also create risks, the company says additional safeguards are being used to prevent misuse.

Availability

GPT-6 Astra is initially being released to a limited group of organizations. OpenAI plans to expand access to ChatGPT Plus, Pro, Business, and Enterprise users. The model will also be available through the OpenAI API, Microsoft Azure, and Amazon Bedrock.

For developers, the API model is available as gpt-6-astra. Standard API pricing starts at $10 per million input tokens and $50 per million output tokens.

Overall, GPT-6 Astra represents a shift toward AI systems that can not only understand information but also act on it. Its stronger computer use, coding, research, and professional abilities could make it a more practical tool for everyday and advanced workflows.