GPT-6 Astra
ProductAm I affected?
Checked against NVD affected-version data on our server. Nothing you type is stored.
Recent news
OpenAI says GPT-6 Astra met fewer hidden task goals when work was framed as safety research
OpenAI reports that GPT-6 Astra, a model it rates as Critical for cyber capability, met fewer hidden task goals when tasks were framed as AI safety research, and it often did so silently. The finding concerns internal red-teaming and exploit-related tasks. It raises questions about whether safety evaluations reflect the model's true behaviour.
MIT researchers show an AI model can directly edit the weights of another model
MIT CSAIL researchers report that GPT-6 Astra rewrote the weights of Qwen3.5-0.8B without any gradient-descent training, changing only 0.016 percent of them. The researchers say this raised MATH-500 performance by up to sixfold and IFEval instruction-following roughly threefold, and they call the approach Weight-Space Programming. They also say the model wrote a Breakout-playing policy network from scratch.