GPT-6 Astra

OpenAI launched GPT-6 Astra on September 3, 2026, its most capable model yet and the first to use what the company calls recurrent depth, a looped reasoning technique that lets the model revisit and refine its read of a problem multiple times rather than reasoning in a single forward pass. Pricing runs $10 per million input tokens, $1 for cached input, $12.50 for cache writes, and $50 per million output tokens, with a 1.05 million token context window and up to 128,000 tokens of output. The approach has drawn scrutiny from outside safety researchers, including groups at Redwood Research, who note that recurrent depth makes the model’s chain of thought harder for outside observers to monitor compared with earlier step-by-step reasoning models.