Our Bureau
San Francisco, CA
OpenAI has launched GPT-5, its new flagship large language model, making it instantly available to all ChatGPT users and API developers. CEO Sam Altman described the upgrade as a “dramatic leap” over previous versions, likening it to the first iPhone with a Retina display — something you wouldn’t want to go back from.
GPT-5 is designed to be smarter, faster, and less prone to inaccuracies than GPT-4. Altman compared the progression of the models to moving from a conversation with a high school student (GPT-3) to a college student (GPT-4), and now to “a PhD-level expert.” It’s integrated into ChatGPT as a single model, with an automatic router that switches to a reasoning version for complex queries.
The model boasts industry-leading coding performance, excelling on benchmarks like SWE-Bench and Aider Polyglot. In demonstrations, it generated a fully functional French-learning website within seconds. OpenAI has also introduced four personality settings — “Cynic,” “Robot,” “Listener,” and “Nerd” — plus chat color customization.
Safety was a key focus, with over 5,000 hours of testing aimed at reducing hallucinations and improving honesty. GPT-5 offers “safe completions” for sensitive queries, balancing helpfulness with security. It also admits more readily when it cannot answer accurately.
For free ChatGPT users, usage caps apply before switching to a “mini” version. Developers can choose from GPT-5, GPT-5 mini, and GPT-5 nano at varying prices. Altman sees the model as a step toward artificial general intelligence, though he admits it still lacks continuous learning capabilities.
“This is clearly a model that is generally intelligent,” Altman said, “but it’s still missing something quite important.”
