OpenAI Releases GPT-6, Claiming Computer Operation Skills Surpass Average Human Users

OpenAI has officially released GPT-6, its next-generation flagship large language model codenamed "Astra." The company said the model surpasses average human users in computer operation capabilities, delivering superior speed and accuracy while autonomously executing multi-step tasks in complex multi-application and multi-system environments. Unlike previous large language models that relied primarily on conversational interaction, GPT-6 adopts an end-to-end autonomous agent architecture, enabling it to parse on-screen visual information in real time, understand graphical user interfaces, and switch between windows, browse web pages, process files, and operate a range of professional productivity software with the proficiency of a skilled operator.

At the application level, GPT-6 can extract data across software programs, auto-fill forms, or generate games, 3D assets, and web architectures with fully runnable code in a single interaction. In terms of underlying capabilities, the model demonstrates a high level of performance in multimodal integrated reasoning, advanced mathematics, and complex code engineering. In research evaluations involving advanced programming and cybersecurity, it showed strong vulnerability analysis and depth of environmental interaction, achieving high performance ratings across multiple professional benchmarks. For high-level system control privileges, OpenAI has introduced stricter multi-layered security reviews and real-time task monitoring mechanisms to ensure the agent strictly adheres to safety boundaries during prolonged complex operations.