GPt-4o is OpenAI’s most powerful AI model on the market, and the company allows both free and paid users to use it. While it’s pretty powerful and insightful as is, OpenAI decided to provide a gift for the tinkerers out there. According to a new report, you can now fine-tune the GPT-4o model to better deliver results to your liking.
Over the years, OpenAI has delivered new features that make ChatGPT much more of a customized chatbot. Recently, the company gave ChatGPT a memory feature. This allows the chatbot to remember certain things about you as you use it. This way, you won’t need to re-educate it on facts about you. It will already have the context it needs when you start a new conversation. This feature is enabled by default.
Users can now fine-tune the GPT-4o model
If you’re a casual ChatGPT user, then you might not really find much use in this feature. However, if you really like to dig into your chatbot, you’ll like it. Users are now able to fine-tune the model in order to better cater the responses to their liking.
After a model is fully trained on data, a person can fine-tune it so that its outputs match a certain style. This is one step to crafting your own personal chatbot. It doesn’t take much to fine-tune a model, and once you get it down, you’ll be able to get the right responses every time.
Starting today, OpenAI is letting users fine-tune GPT-4o to deliver the responses they want. Users will have access to 1 million free training tokens to fine-tune the model. If you want to fine-tune GPT-4o, then you’re going to want to act fast, as they are only free until September 23rd. After that point, every 1 million input tokens will cost $25. Also, every 1 million output tokens will cost $15.
If you’re going to use this for business purposes, OpenAI ensures that people who use it will have full ownership of their business data. The only issue is that the company has seen some rather bad data breaches in the past, so people might be on the fence about putting their data in OpenAI’s hands.
Also, OpenAI is monitoring how people are fine-tuning the GPT-4o model. This is to make sure that no one steers it to break the company’s usage policies.