OpenAI launches Astra, a powerful and controversial AI model

OpenAI released Astra on Thursday, calling it the company’s most powerful and capable AI model yet. The company said Astra marks “a new frontier on computer and browser use” and handles tasks with unmatched “speed, accuracy, and safety.”

Astra is initially available to OpenAI customers using Daybreak, the company’s cybersecurity program. Over the next week, it will also reach paid Pro, Plus, Enterprise, and Business accounts, as well as the API.

OpenAI highlighted Astra’s cybersecurity and coding abilities. The company said it tested the model against security benchmarks and that it can identify and develop zero-day exploits, which can help defenders find and patch weaknesses. OpenAI also called Astra its “best model for software engineering to date.” The company said benchmark results placed Astra ahead of OpenAI’s Sol and Anthropic’s Fable in tasks including bug finding, terminal execution, and codebase queries.

The model has also drawn controversy over opaque recurrence, a reasoning technique that can obscure chain of thought and make monitoring more difficult. OpenAI chief scientist Jakub Pachocki said increasing model capabilities were making monitorability more challenging.

OpenAI president Greg Brockman said he personally believed “we’re there” regarding artificial general intelligence, while noting that AGI is now a mission or spiritual concept rather than a contractual trigger.