OpenAI's GPT-6 Astra Unveiled with Enhanced Reasoning and New API Primitives, Reaching Critical Cybersecurity Threshold

OpenAI has launched GPT-6 Astra, its most advanced AI model, capable of performing complex computer-based tasks with increased efficiency. The model is notable for being the first to reach the Critical cybersecurity threshold, though it surfaces fewer reasoning steps, which could impact oversight. New API primitives, including the ability to change reasoning effort without invalidating the prompt cache, were also introduced. OpenAI President Greg Brockman suggested the company might have reached AGI.