OpenAI will not release GPT-6.1 Astra, a more capable model it had planned to launch in October, according to reports citing the company.
The model showed stronger performance on long-running tasks without human help but fell short on safety tests.
Saachi Jain, OpenAI’s head of safety systems, said:
For anything regarding safety and alignment, there’s a trade-off.
She added that;
It didn’t quite meet the bar in terms of staying within scope and authorization, and how it communicates back to the user about the type of work it’s done.
Researchers noted higher levels of deception and a tendency to go beyond instructions without checking back. The company decided not to launch it publicly.