Upgrade to New Models: GPT-3 to GPT-3.5 Turbo and GPT-4
Starting from January 4, 2024, applications utilizing the stable model names for base GPT-3 models (ada, babbage, curie, davinci) will be automatically upgraded to the new models. The new models, namely ada-002, babbage-002, curie-002, and davinci-002, will be widely accessible for early testing in the upcoming weeks by specifying these model names in API calls.
Upgrading from Older Completion Models
Developers who currently use other older completion models, for example, text-davinci-003, will need to manually upgrade their integration by January 4, 2024. This can be done by specifying the model parameter of their API requests as gpt-3.5-turbo-instruct. The gpt-3.5-turbo-instruct is built like InstructGPT and trains similar to text-davinci-003. It can be easily integrated into the Completions API and will be available for early testing soon.
Migrating from Fine-Tuned Models
For developers who wish to continue using their fine-tuned models beyond January 4, 2024, they will need to fine-tune replacements on the new base GPT-3 models (ada-002, babbage-002, curie-002, davinci-002) or the newer models (gpt-3.5-turbo, gpt-4). Users with previously fine-tuned older models will receive priority access to GPT-3.5 Turbo and GPT-4 fine-tuning when this feature becomes available later this year. We understand that migrating from models fine-tuned on personal data can be challenging, and we are committed to providing support to ensure a smooth transition.
Details regarding the new completion models’ readiness for early testing will be shared with developers who have recently used the older models. Stay tuned for more information in the coming weeks.