OpenAI on Thursday unveiled GPT-5, calling it its “smartest, fastest, most useful” large-language model and an important step toward artificial general intelligence. The model is available immediately in ChatGPT and via an API, replacing GPT-4o and earlier versions after more than two years of development setbacks and internal testing. GPT-5 comes in four variants—standard, mini, nano and a deeper-reasoning “thinking” mode—and is governed by a real-time router that decides when to invoke extended reasoning. OpenAI says the system cuts factual errors by about 45 % compared with GPT-4o, introduces so-called “safe completions” to answer sensitive queries, and shows marked gains in coding, mathematics, writing and health-related advice. Access is being extended to the entire ChatGPT user base. Free users receive up to 10 GPT-5 messages every five hours and one thinking-mode request per day before reverting to a lighter model, while Plus subscribers get 80 messages every three hours and Pro or Team customers enjoy unlimited usage and early access to GPT-5 Pro. Enterprise and education clients will follow next week. Microsoft simultaneously activated GPT-5 across its Copilot portfolio, bringing the model to Microsoft 365 Copilot, GitHub Copilot and Azure AI Foundry at no additional cost, underscoring the companies’ deepening partnership. CEO Sam Altman described the launch as a "significant step" in OpenAI’s race with Google, Anthropic and xAI to dominate the fast-moving generative-AI market.