OpenAI launches GPT-3.5 Turbo fine-tuning feature to create an exclusive ChatGPT chat bot

Fine-tuning for GPT-3.5 Turbo is now available, and fine-tuning for GPT-4 will be available this fall. This update gives developers the ability to customize models that are better suited for their tasks and run those models at scale. Initial tests have shown that the fine-tuned version of GPT-3.5 Turbo can match or even surpass the base-level capabilities of GPT-4 for some narrow tasks. As with all of our APIs, the data passed to the fine-tuning API is proprietary and is not used by OpenAI or any other organization to train other models.

Examples of how to use fine-tuning
Since the release of GPT-3.5 Turbo, developers and companies have been asking for the ability to customize the model to create a unique and differentiated experience for their users. With the release of this version, developers now have the ability to perform fine-tuning of the model to make it more effective.

In private beta testing, customers were able to significantly improve the model's performance in the most common use cases, such as:

Improved manageability: Fine-tuning allows businesses to make the model follow instructions better, such as making outputs more concise or always responding in a given language. For example, developers can use fine-tuning to ensure that the model always responds in German when asked to use that language.

Robust output formatting: Fine-tuning improves the model's ability to consistently format responses, which is important for applications that require a specific response format, such as for code completion or composing API calls. Developers can use fine-tuning to more reliably convert custom prompts into high-quality JSON snippets that can be used in their own systems.
Custom Tone: Fine-tuning is a great way to improve the quality characteristics of a model's output, such as its tone, to better match a company's brand voice. Companies with a recognizable brand identity can use model fine-tuning to better match their tone.

Fine-tuning not only improves performance, but also reduces the number of prompts while providing similar performance. Fine-tuning with the GPT-3.5-Turbo also allows us to process 4k tokens - twice as many as our previous fine-tuned models. The first testers reduced the size of the prompts to 90% by fine-tuning the instructions in the model itself, speeding up each API call and reducing costs.

Fine-tuning is most effective when combined with other techniques such as designing tooltips, information retrieval, and feature recall. Check out our guide to fine-tuning to learn more. Support for fine-tuning with function calls and gpt-3.5-turbo-16k will be implemented this fall.