Currently, all GPTs are using GPT-4, which works ok, but not sexy enough to be a great commercial product. For better results, my team would like to have the option to use a fine-tuned GPT-4/GPT3.5 turbo as the base model for our custom GPT.
Currently, we do not want to go through Assistant API for this purpose; instead, we like the current flow of using GPT builder, which is hosting (and has a frontend ) by OpenAI’s. @logankilpatrick
Just a bit curious why you do not like to use Assistant API, which is supposed to be more powerful and is higly possible to apply for commercial purpose and you could create products hosting by your own while not by OpenAI. Are there some considerations about the cost?
1 Like
I am looking for confirmation if the model used by GPT builder is it really the available gpt-4 model? it looks to behave much better for the same prompts for chats than using the straight api-s (or even the assistant). Is this not a custom finetuned model?