OpenAI has finally released GPT-5.
The company is calling it a unified system that can decide how hard to think before giving you an answer. Sam Altman framed GPT-5 as the moment where you can “ask a legitimate expert, a PhD-level expert, anything” and get a reliable answer.
The new model is gradually rolling out to all 700 million ChatGPT users — yes, that includes even those without paid subscriptions.
After watching the one-hour announcement, the question is simple: Does GPT-5 really deliver a meaningful jump from previous models, or was it overhyped?
In this article, I’ll discuss the following topics:
What is GPT-5?
What’s new in GPT-5
Performance and evaluation data
My personal observations on the launch
Availability and API pricing
Final thoughts
Let’s get started.