OpenAI on Thursday unveiled GPT-6 Astra, a new artificial intelligence (AI) model that the tech company says features major improvements in computer use, coding, scientific research and professional work.

The San Francisco-based company said Astra is rolling out starting Thursday to enterprise customers with Daybreak access. In the coming days, it will also become available to ChatGPT Plus, Pro, Business and Enterprise users, as well as through the OpenAI API and Amazon Web Services.

"GPT-6 Astra is built on advances across pre-training, reinforcement learning, and alignment, bringing together years of research and big bets," OpenAI said. "Astra is state-of-the-art on computer use, browser use, software engineering, cybersecurity, science, and professional work."

According to OpenAI, the new model can carry out "tedious" computer-based tasks such as filling out online forms, updating customer records, organizing calendars and conducting research. The company also said Astra can create documents, spreadsheets and presentations. Such computer-use capabilities reflect a broader industry push toward agentic AI — models that can complete multi-step tasks on a user's behalf rather than only answering questions — an area where OpenAI competes with rivals including Google, Anthropic and other major AI labs.

"GPT-6 Astra is the best model for software engineering to date," the company added.

OpenAI also said that Astra represents a significant jump in cybersecurity capabilities.

"As we discussed in our safety update, Astra is a significant jump in cyber capabilities and meets the Critical threshold in cybersecurity under our Preparedness Framework," OpenAI said. Under that framework, OpenAI evaluates frontier models for potentially dangerous capabilities before release, with defined risk thresholds that determine whether and how a model can be deployed.

The company also touted Astra's performance in math and science, calling it a "major advance for scientific discovery, mathematics, and health."

"Astra is our most aligned model. Astra excels at exercising care, respecting task boundaries, and communicating transparently," the company said. "This work is the latest product of our long-running research program focused on training models that remain aligned with human intent from start to finish."

The company has faced heightened scrutiny over AI safety after OpenAI-built agents breached their testing environment and accessed the open-source platform Hugging Face. OpenAI said Astra was not involved in that incident. The episode has drawn attention to how AI developers contain and monitor autonomous agents as their capabilities grow.

Reuters contributed to this report.