OpenAI Launches GPT-5, A Smarter, Faster AI With A Unified Agent System
OpenAI has revealed that the model is becoming available to all users, replacing the previous GPT-4o in all tiers of ChatGPT. However, Plus subscribers will gain access to additional functionality and daily usage credits, while Pro subscribers will get unlimited access to GPT-5 Pro, the most powerful variant of GPT-5. OpenAI describes GPT-5 Pro as "a version with extended reasoning for even more comprehensive and accurate answers."
The AI race is really getting interesting. Just yesterday, Google announced that its first coding agent, Jules, graduated from the beta phase and is now available to the public. Now, there is GPT-5 which OpenAI has described as its "strongest coding model to date." Well, in the livestream debuting the new model, OpenAI CEO Sam Altman said it is "well beyond vibe coding," and capable of "very complex software engineering tasks in practice."
OpenAI showcased the programming capabilities of GPT-5 by publishing a set of five prompts and the resulting apps generated by GPT-5. The results are impressive; below, you can see a screenshot of the apps coded by GPT-5 from a single prompt. It's a simple runner game, but it's fully playable and even has sounds.
There are five application demos apparently created from a single prompt by GPT-5.
It's difficult to explain how impressive this is. With one prompt, this model has crafted program code, simple graphics, and sounds, all for an interactive application that really works. You can try the game for yourself, as well as four other GPT-5-coded projects, at OpenAI's blog linked below.
Aside from coding, GPT-5 also excels in writing, math, and health thanks to being a fully unified model, meaning that users don't have to manually select thinking or research modes. OpenAI says that GPT-5 is specifically designed to answer health-related questions with high accuracy; Sam Altman described the new model as akin to having a "PhD-level expert in your pocket, on any topic." Of course, it also warned that "ChatGPT does not replace a medical professional."
OpenAI has also published some benchmarks to help users understand how smart GPT-5 is, although it only compares its models to previous OpenAI work instead of competitor models. You can check out these diagrammatic illustrations of GPT's new capabilities in the blog above, but you might do better to just head to the ChatGPT site and test the model for yourself, as GPT-5 either is or soon will be the new default chatbot for every signed-in user.
