Yesterday, ChatGPT officially announced on social network X in a low-key manner that the AI tool has been using the new GPT-4o model since last week. ChatGPT stated, "Since last week, a new GPT-4o model has appeared in ChatGPT. We hope you all enjoy it; if you haven't experienced it yet, check it out! We believe you'll like it."
Although the ChatGPT application account on X did not immediately provide more information after the post was published, insiders from OpenAI revealed that the new model was updated based on user feedback.
In subsequent release notes, OpenAI explained that they introduced updates to the GPT-4o model and, through experiments and user feedback, found that users preferred this new version. Although it is not a completely new model, they are working on finding better ways to measure and communicate the improvements in the model's behavior.
Many users found that the new model seems to be able to conduct more detailed step-by-step reasoning and provide more detailed explanations. However, a spokesperson for OpenAI said that the reasoning process of the model has not changed, and ChatGPT primarily responds to specific user prompts when describing its reasoning. Before the official announcement, many users had already noticed that ChatGPT's performance seemed to have improved.
Additionally, users reported that the image generation capabilities of GPT-4o in ChatGPT have also been activated. Although the previous GPT-4 model already had the ability to generate images, this update allows GPT-4o to generate higher quality images more efficiently. The new model can not only process text but also convert pixels into tokens, achieving more accurate and realistic image generation.
Although many users welcomed this update, some criticized it, believing that OpenAI should explain more clearly the changes in the model's user experience and behavior. Some users even felt that the update was not significant enough and seemed superficial.
When asked whether there are different versions of GPT-4o in ChatGPT and the API, an OpenAI spokesperson said that they often make small improvements in ChatGPT and the API. They will release the best version for developers' needs in the API.
OpenAI confirmed on their official developer account that the new model has been rolled out in the API, and developers can use "chatgpt-4o-latest" to test the latest improvements. Details about this update are also actively being addressed by the OpenAI team in response to user and developer questions.
Key Points
🌟 New GPT-4o model launched, modified based on user feedback.
🤖 Users find the model performs better, capable of more detailed step-by-step reasoning.
🖼️ New model supports image generation, enhancing quality and efficiency.