Ive tried raising and lowering the temperature and top_p parameters, Ive tried experimenting with different models, both the gpt-4-turbo-2024-04-09 used by the chat window, and also tried model gpt-4o-2024-08-06.
I still get the same output in the api, but this is different than the chat window.
How do I configure the api to return the same output as the chat window?
Welcome. If you search the forums, you’ll find a wealth of information.
Hope you stick around.
2 Likes
ChatGPT is a developed product, using a different model than the API and with proprietary messages and tools placed that must, by the very nature of the technology, affect the output, besides the fact that it is designed to be creative in every new response.
Compare sending the exact same ChatGPT input message into two different new chat sessions. ChatGPT will be different than ChatGPT.
The API is for unique development, not a different way to talk to the same consumer web product that OpenAI sells with subscription.
1 Like