Which Model is the Best for Writing?

I have a program that writes essays. Does anyone know which model will give the highest-quality result?
I assume it is either gpt-4 or gpt-4-1106-preview? If so, which of the two gives the highest quality?

Define quality, please. Do you mean how accurate are facts, how smooth that text is or if a reader sees in seconds it is creation of a GPT?

1 Like

What most people would do in your situation is try each of the models several times and decide which they prefer.

Without knowing all the details of your system or the types of essays your trying to generate it’s literally impossible for anyone to answer this question for you.

Also much depends on how construct your prompt. As others have said, it’s really down to testing. Depending on your specific needs, you might also consider a fine-tuned model.

1 Like

I am defining quality by “how accurate are facts” and “how smooth is the text”.

Does the community have a consensus on the difference between the two models? For example, does one get their facts wrong more? Does one write more robotically? You’re right, I will test myself. But I also would like to know if people already have a consensus on the difference in writing quality between these two models.

Thanks. I have never fine-tuned a model, but I can give it a try.

Nope - alot about LLM usage is contextualised to one’s use case.

1 Like