When is fine-tuning available for the gpt-3.5-turbo?


When is fine-tuning available for the gpt-3.5-turbo?

If we fine-tune our data on GPT-3, we will have to fine-tune again when the newer version (3.5 or 4.0) is available? This will cost us more money so we want to find out when the new version is coming out…

I wonder if we have to fine tune every time new version is available…



Is finetuning out for GPT-3 already ? As far as I remember, you could only finetune davinci base model.

In this blog post OpenAI talks about fine-tuning ChatGPT, but it’s not the same as the current fine-tunes of the base models. No timeline.

Fine-tuning is not available yet for GPT-3.5-turbo and there is no official announcement on when it will be.

In this blog: How should AI systems behave, and who should decide?
It only discussed that who we can fine-tune ChatGPT and make it better.
It’s not saying that we can fine-tune ChatGPT into other commercial use.

After playing with gpt-3.5 a good amount, it is extremely fine tuned to be a chatbot, which has limited applications in a lot of other areas. Fine tuning would be extremely helpful, but I have a feeling it is likely limited to those with the deepest pockets for now.

I don’t understand who could use this to provide chat support, since the bot is inventing information that was not provided. I was playing around in the playground and fed it with certain data such as plan names, benefits and even clarifying that the price should be consulted, yet it is still inventing values, among other things.

For translations, for example, it was useful for me. It was difficult to find a prompt that works well, but for chat, I don’t see a way to use it without having fine-tuning tools.

1 Like

There are a multitude of approaches with providing it the necessary training to be a chat support VA. Semantic Searching, pairing with existing IVRs like dialog flow cx, Intents and Entities, custom crms, I mean honestly this is so cutting edge that Im sure there are ways that no one has thought of.