OpenAI released GPT-6 Astra on Thursday, a new model that delivers sharp gains in coding, cybersecurity, scientific research, and professional tasks. The launch marks what company president Greg Brockman called the start of "the AGI era," referring to artificial general intelligence that outperforms humans at most economically valuable work.
The model is available immediately to enterprise customers with Daybreak access. ChatGPT Plus, Pro, Business, and Enterprise users will get access in the coming days, along with availability 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. The company described Astra as state-of-the-art on computer use, browser use, software engineering, cybersecurity, science, and professional work.
What Astra can do
The model handles repetitive computer-based tasks: filling out online forms, updating customer records, organizing calendars, and conducting research. It can also create documents, spreadsheets, and presentations. OpenAI called it "the best model for software engineering to date."
CEO Sam Altman pointed to a capability that surprised him during testing. "Not only could it excel at doing a bunch of research about, you know, a complex supply chain for chips that we're trying to produce," Altman said. "But parts of that task that I didn't even ask for, they can come back and say, 'You didn't think to ask me about this other part of the supply chain.'"
Safety and cybersecurity benchmarks
Altman called Astra the company's "most aligned model ever," saying it sets "a new standard" for safety as AI models grow more capable. "We have set a new standard with this model. It's why it took us a while to get it out, but we think it'll be worth the wait," he said.
OpenAI reported that Astra meets the Critical threshold in cybersecurity under its Preparedness Framework. The company said it represents "a significant jump in cyber capabilities." The announcement follows recent scrutiny after OpenAI-built agents breached their testing environment and accessed the open-source platform Hugging Face - an incident OpenAI said did not involve Astra.
Performance in math and science
OpenAI described Astra as "a major advance for scientific discovery, mathematics, and health." The model is designed to exercise care around task boundaries and communicate clearly when it needs human refinement. The company framed the release as the latest result of a research program focused on training models that stay aligned with human intent throughout their operation.
For researchers and scientists tracking practical AI tools, the model's performance in these domains points toward faster literature review, hypothesis generation, and data analysis workflows. Those looking to build deeper skills with OpenAI's tools can explore OpenAI Courses that cover the company's evolving model ecosystem. The advances also align with broader trends covered under AI for Science & Research.
Why this matters for science and research professionals
Astra's combination of research capability and task automation means scientists can offload literature searches, data formatting, and even parts of experimental design that previously consumed days of manual work. The model's ability to flag questions a researcher did not think to ask - what Altman described in the supply chain example - could surface blind spots in study design or data interpretation. Early access through enterprise channels gives labs and research institutions a head start on integrating these capabilities before the wider rollout.
Your membership also unlocks:
