Enhance/train GPT-4 Model

I would like to use GPT-4 in our company with internal knowledge. Is it possible to expand/train the existing gpt-4 model with internal knowledge. I know the max. number of tokens per request are 32000 - but in companies there is much more knowledge than 32000 tokens.

Is there a way to extend this model or will it be a feature for future? I think - this requirement will be one for a lot of companies. Maybe it will be possible like train your own model - like it is possible in stable diffusion.

We are a company developing accounting programs. We would use GPT-4 for customer-support “support”.


Semantic search using embeddings, so that you can include the most relevant context from your knowledge base in your prompt, is the generally accepted way to proceed in cases like this.

1 Like

Thank you, I’ll try to train our internal knowledge with embeddings and see if I can get what I want.