OpenAI, a research company focused on artificial intelligence, has announced new features and expanded programs to empower developers in creating custom AI models.
What is fine-tuning? Fine-tuning is a technique used to improve the performance of a large language model (LLM) on a specific task. In other words, fine-tuning refines an LLM’s abilities for a particular use case.
OpenAI’s Fine-Tuning API
OpenAI offers a self-serve fine-tuning API (Application Programming Interface) that allows developers to customise large language models like GPT-3.5. This Application Programming Interface (API) has been used by organisations such as Indeed to improve their workflow and performance.

“For example,” OpenAI revealed, “Indeed, a global job matching and hiring platform, wants to simplify the hiring process. As part of this, Indeed launched a feature that sends personalized recommendations to job seekers, highlighting relevant jobs based on their skills, experience, and preferences. They fine-tuned GPT-3.5 Turbo to generate higher quality and more accurate explanations. As a result, Indeed was able to improve cost and latency by reducing the number of tokens in prompt by 80%. This let them scale from less than one million messages to job seekers per month to roughly 20 million.”
Benefits of Fine-Tuning
A fine-tuned model gains a deeper understanding of specialised fields like law or medicine, it can improve its accuracy in tackling related tasks, and it also delivers results faster and cheaper by requiring less prompting.
New Fine-Tuning API Features
OpenAI is introducing several new features to enhance the developer experience:
- Epoch-based Checkpoint Creation: Automatically produce one full fine-tuned model checkpoint during each training epoch, which reduces the need for subsequent retraining, especially in the cases of overfitting.
- Comparative Playground: A new side-by-side Playground UI for comparing model quality and performance, allowing human evaluation of the outputs of multiple models or fine-tune snapshots against a single prompt.
- Third-party Integration: Support for integrations with third-party platforms (starting with Weights and Biases this week) to let developers share detailed fine-tuning data to the rest of their stack.
- Comprehensive Validation Metrics: The ability to compute metrics like loss and accuracy over the entire validation dataset instead of a sampled batch, providing better insight on model quality.
- Hyperparameter Configuration: The ability to configure available hyperparameters from the Dashboard (rather than only through the API or SDK).
- Fine-Tuning Dashboard Improvements: Including the ability to configure hyperparameters, view more detailed training metrics, and rerun jobs from previous configurations.
OpenAI further revealed that it offers a Custom Models Program for organisations that require a deeper level of customisation. This program includes two options:
- Assisted Fine-Tuning: OpenAI collaborates with developers to leverage advanced techniques beyond the API, maximising model performance for their specific needs.
-
- Example: SK Telecom, a South Korean telecoms provider, partnered with OpenAI to fine-tune a model for improved customer service interactions in Korean. This resulted in a significant boost in conversation summarisation, intent recognition, and customer satisfaction scores.
- Custom-Trained Models: For highly specialised tasks, organisations can build entirely new models from scratch using vast amounts of domain-specific data.
- Example: Harvey, a legal AI tool provider, collaborated with OpenAI to create a custom model for legal case law. This model achieved an 83% increase in factual responses and was overwhelmingly preferred by lawyers compared to the base GPT-4 model.
“We believe that in the future, the vast majority of organisations will develop customised models that are personalized to their industry, business, or use case. With a variety of techniques available to build a custom model, organisations of all sizes can develop personalized models to realise more meaningful, specific impact from their AI implementations. The key is to clearly scope the use case, design and implement evaluation systems, choose the right techniques, and be prepared to iterate over time for the model to reach optimal performance,” the technology company said.