On August 7, 2025, OpenAI released GPT-5. In its announcement “Introducing GPT-5,” OpenAI described it as a unified system that “knows when to respond quickly and when to think longer,” automatically deciding whether a question needs a fast answer or extended step-by-step reasoning - folding the separate reasoning-model track (the o-series) back into the main GPT line. GPT-5 was made available to all ChatGPT users, with higher usage tiers for paying subscribers and a more capable GPT-5 pro variant.
OpenAI reported state-of-the-art results across coding, math, and multimodal tasks, including 74.9 percent on SWE-bench Verified for real-world coding and 94.6 percent on the AIME 2025 math competition without tools.
GPT-5 mattered as the headline release in a year of rapid frontier-model competition, and as a design statement: rather than asking users to pick between a fast model and a slow reasoning model, OpenAI merged them so the system chooses how hard to think. That single-model, auto-routing approach became a template other providers followed.
(Sourcing note: OpenAI’s announcement page is the Tier 1 primary; its date and claims were confirmed through multiple independent quotations of that page, as openai.com blocks automated fetching.)