OpenAI announces GPT-3.5 Turbo fine-tuning and other API updates
OpenAI reported that fine-tuned GPT-3.5 Turbo could match or beat base GPT-4 on some narrow tasks, and said fine-tuning for GPT-4 itself would follow that autumn.
- Models & capabilities
- Minor
OpenAI opened up fine-tuning for GPT-3.5 Turbo through its API, letting developers train custom versions of the model on their own labelled data rather than relying only on prompt engineering with the base model. The company said data submitted for fine-tuning remained the customer’s own and would not be used to train other OpenAI models.
OpenAI reported that on narrow, well-defined tasks, a fine-tuned GPT-3.5 Turbo model could match or even exceed the performance of the much larger base GPT-4 model — a claim aimed at developers weighing cost and latency against capability, since GPT-3.5 Turbo was substantially cheaper to run. The announcement also said fine-tuning for GPT-4 itself was planned for later that year.
The update was one of a steady stream of incremental API changes OpenAI shipped through 2023 as it built out a developer platform alongside its consumer ChatGPT product, and it reflected a broader industry shift toward letting customers customise general-purpose models for specific applications rather than always calling a single shared model.