OpenAI unveils GPT-6 Astra

GPT-6 Astra boosts science, math, coding, and cybersecurity with top benchmark scores See what the rollout means for ChatGPT and API users as access expands soon

OpenAI has introduced GPT6 Astra, describing it as a major new model with stronger performance across science, math, coding, cybersecurity, and computer use. The company and its president, Greg Brockman, framed the release as a significant step forward in AI capability. According to the article, Astra posted standout benchmark results, including very high scores on ARCAGI3 and FrontierMath T4, along with a perfect score on ExploitBench for cybersecurity. OpenAI also said the model is more efficient in how it uses tokens for tasks, though its API pricing is higher than the previous model. Access will roll out gradually, starting with a small group of organizations before reaching paid ChatGPT users and API customers in the following days. The article notes that the model’s realworld impact will become clearer once more users can try it.