OpenAI said GPT-4o fine tuning, which will enable you to customize the company's best model with proprietary data sets, is available to developers across all paid usage tiers.
In a post, OpenAI said fine tuning for GPT-4o was one of the most requested features from developers. OpenAI said that it will offer 1 million training tokens per day for free through Sept. 23. For GPT-4o mini, OpenAI is offering 2 million training tokens per day for free through Sept. 23.
Fine tuning will enable developers to customize GPT-4o for various use cases. Enterprises will be interested in fine tuning for GPT-4o since it can lower costs.
The models available for fine tuning include:
- gpt-4o-2024-08-06
- gpt-4o-mini-2024-07-18
- gpt-4-0613
- gpt-3.5-turbo-0125
- gpt-3.5-turbo-1106
- gpt-3.5-turbo-0613
- babbage-002
- davinci-002
GPT-4o fine tuning will cost $25 per million tokens and inference is $3.75 per million input tokens and $15 per million output tokens.
OpenAI highlighted examples of companies using GPT-4o for fine tuning and noted that enterprises have full ownership of their business data and all inputs and outputs. Common use cases for fine tuning include setting style, tone and format, improving reliability, correcting failures, handling edge use cases and performing a new skill that's hard to articulate in a prompt.
