GPT-5 is the model OpenAI released on August 7, 2025, folding its previous split between fast chat models (GPT-4o) and slow reasoning models (the o-series) into one system that decides on its own how much internal reasoning a given prompt needs. Rather than making users pick a model, ChatGPT and the API route each request to a lighter or heavier internal mode automatically, cutting the number of consumer-facing choices while still letting API customers set an explicit reasoning effort. It ships with a 400K token context window (272K input, 128K output), API pricing of $1.25 per million input tokens and $10 per million output tokens, and scored around 90 on MMLU alongside strong marks on coding and agentic benchmarks at launch. OpenAI framed it as a jump in reliability and reduced hallucination rate rather than a dramatic leap in raw capability, positioning it against Google’s Gemini 2.5 Pro and Anthropic’s Claude models of the same era. By mid-2026 OpenAI had already iterated past it with the GPT-5.x line, including the GPT-5.6 family, but GPT-5 remains available in the API and ChatGPT.