GPT-5 Nears Human Parity

OpenAI is testing GPT-5 to assess how well it performs against humans in economically valuable roles. GPT-5 demonstrates advancements in coding, mathematics, health, and visual perception. It matches or exceeds human expert performance in roughly half of 40 professions. The model integrates a fast core, a reasoning component, and a real-time router for optimal query handling. GPT-5 showcases improvements in complex front-end generation and debugging, producing visually appealing applications with enhanced typography and spacing.

GPT-5 exhibits enhanced capabilities in writing, coding and health. It can now handle structurally complex writing tasks and supports creative expression. The model demonstrates improved decision-making for tool selection, reducing errors in long-term tasks. GPT-5 is also less prone to factual errors and hallucinations compared to previous models. New features include preset personalities for tailored interactions. GPT-5 is available to all ChatGPT users, with higher usage limits for Plus subscribers and access to GPT-5 Pro for Pro subscribers.