OpenAI Launches Fine-Tuning Feature for GPT-4o

OpenAI launched fine-tuning for GPT-4o, saying this capability allows developers to increase performance and accuracy for their applications.

The feature launched Tuesday (April 20) and is available to all developers on all paid usage tiers, the company said in a Tuesday blog post.

“Developers can now fine-tune GPT-4o with custom datasets to get higher performance at a lower cost for their specific use cases,” OpenAI said in the post. “Fine-tuning enables the model to customize structure and tone of responses or to follow complex domain-specific instructions. Developers can already produce strong results for their applications with as little as a few dozen examples in their training dataset.”

OpenAI tested the fine-tuning feature “with a handful” of partners before launching it, according to the post.

In one case, it found that a company that offers an AI software engineering assistant used fine-tuning to enable the model to learn to respond in a specific way and output in specific formats, the post said.

“Fine-tuned models remain entirely under your control, with full ownership of your business data, including all inputs and outputs,” OpenAI said in the post. “This ensures your data is never shared or used to train other models.”

Domain-specific training is key to the use of generative AI in the workplace, PYMNTS reported in November 2023. Some observers believe that vertically oriented AI models, or large language models and GPT systems trained for industry-specific use cases on validated and audited datasets that are regularly retrained and updated per sector-specific guidelines, and can be fine-tuned by organizations using their own data, offer enterprises the safest, most productive way forward.

Meanwhile, OpenAI said in July that it began rolling out an advanced voice feature for its ChatGPT platform. The feature, which uses the GPT-4o model, offers hyper-realistic audio responses and enables users to have real-time, delay-free conversations with ChatGPT and even interrupt it mid-sentence.

The alpha version of Advanced Voice Mode was released to a select group of ChatGPT Plus subscribers, with the company planning to begin a broader rollout to all premium users in the fall.

For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.