OpenAI announced on Wednesday that the performance-optimized version of ChatGPT's newest large-scale language model, GPT-4o, will enable users to experience more natural, engaging, and personalized writing content.
First released in May 2024, GPT-4o is twice as powerful as its predecessor, GPT-4-Turbo, and uses half the computing resources. The model delivers outstanding results in speech, multilingual and vision tasks, and provides faster response times, making it particularly well-suited for on-the-fly translation and dialog applications.
Notably, GPT-4o has advanced reasoning capabilities to understand the user's tone of voice, intent, and emotions and respond accordingly. This technology provides the ChatGPT of advanced speech patterns provide critical support.
In terms of usage restrictions, free users will only be able to use GPT-4o a limited number of times, after which they will move to the smaller GPT-4o-mini model, while Plus, Teams, and Enterprise users will enjoy approximately five times the amount of usage.
ground Artificial Analysis The data shows that GPT-4o-mini scored 82% in the MMLU inference test, outperforming the Google Gemini 1.5 Flash 3%, and Anthropic Claude 3 Haiku 7% results. openAI says this is the most cost-effective miniaturized model available.
How to experience the latest version of GPT-4o
After upgrading the model number to gpt-4o-2024-11-20, login to ChatGPT, there are chances to use: gpt-4o-2024-11-20
Or access it directly using the API.
How is it verified?
Enter "Hello, what is the cutoff time for data in your training set?" to see the knowledge cutoff date.
If it's 2021/2022, then the ChatGPT or API you're using has been "downgraded", which may have something to do with your network environment.
When 2023 is shown, it is normal, when November 2024 is shown, the upgraded gpt-4o-2024-11-20 is used.
English 
简体中文 
日本語 
Deutsch (Sie)