Customers will have to upload data to OpenAI's servers if they want to fine-tune GPT-4o.

Currently, startups are fiercely competing in the enterprise AI product segment, while businesses face the problem of finding profits from AI.

The customization capabilities of an AI model include generating additional information about specific task types or subject areas, for example, a skateboard manufacturing company could tweak the GPT-4o AI model to create a chatbot capable of answering questions about wheels or skateboard maintenance procedures.

OpenAI never opened up this new feature with GPT-4o or its predecessor, GPT-4. However, the company has allowed users to customize many other AI models, such as GPT-4o mini (a smaller, cheaper version of GPT-4o).

Providing the ability to customize AI models will allow customers to work directly with the company instead of using a middleman or inferior products, said OpenAI's API product manager.

“We've focused a lot on lowering the barrier for customers,” the executive said.

However, it is worth noting that to fine-tune the model, customers must upload their data to OpenAI's servers and the training process will take about 1-2 hours.

OpenAI currently only offers customers the ability to self-train using text data, not expanding to image data or other content.