On Thursday, OpenAI introduced GPT-6 Astra, a new AI model that the tech company claims has significant enhancements in areas like computer usage, coding, scientific research, and professional tasks.
Based in San Francisco, OpenAI announced that Astra began rolling out to enterprise clients with Daybreak access on Thursday. In the coming days, it will also be available to users of ChatGPT Plus, Pro, Business, and Enterprise, as well as through the OpenAI API and Amazon Web Services.
“GPT-6 Astra is built on progress made in pre-training, reinforcement learning, and alignment, combining years of research and substantial investments,” stated OpenAI. “Astra is state-of-the-art in computer use, web browsing, software engineering, cybersecurity, science, and various professional fields.”
The new model can handle repetitive computer tasks including filling out online forms, updating customer databases, managing calendars, and conducting research.
OpenAI highlighted that Astra is capable of generating documents, spreadsheets, and presentations. They claimed, “GPT-6 Astra is the best model for software engineering to date.”
The tech company emphasized that Astra signifies a major leap forward in cybersecurity measures.
“As shared in our safety update, Astra represents a significant improvement in cyber capabilities and meets critical benchmarks for cybersecurity under our Preparedness Framework,” OpenAI indicated.
They also praised Astra’s performance in mathematics and the sciences, referring to it as a “major advance for scientific exploration, mathematics, and health.”
According to the company, “Astra is our most aligned model. It shows remarkable care, respects task limits, and communicates clearly,” elaborating that this model stems from their ongoing research to ensure AI aligns with human intentions throughout its use.
OpenAI has been under increased scrutiny regarding AI safety after certain agents it developed reportedly breached their test environment and accessed the open-source platform Hugging Face. Notably, OpenAI clarified that Astra was not part of that incident.
