OpenAI Just Released GPT-6 Astra and it scored 100% on cybersecurity OpenAI's GPT-6 is being called a step toward AGI, and the test scores explain why. On completely novel problems with no instructions, it scored 99.9%. On one of the hardest AI math tests, 97.6%. On cybersecurity, 100%, capable of finding unknown software flaws and turning them into working exploits, enough to land it in the highest risk safety category. But that headline 99.9% wasn't under standard test conditions. Under typical setup, it dropped to 62.7%, still strong, just a different story. Share this with someone debating AGI. Save it, the numbers matter here. #GPT6 #OpenAI #AGI #AI #anik