⭐️ OpenAI Model Deprecation Guide

⭐️ OpenAI Model Deprecation Guide

In two days (i.e. Jan 4), OpenAI will retire 33 models, including GPT-3 (text-davinci-003) and various others. This is OpenAI's biggest model deprecation so far. Here's what you need to know:


GPT-3 Model Retirement

The text-davinci-003 model (commonly known as GPT-3) will be unavailable from Jan 4.

→ You must manually transition to the replacement model, gpt-3.5-turbo-instruct .

/fine-tunes Endpoint Retirement

The ada, babbage, curie, and davinci models on the /fine-tunes endpoint will be retired on Jan 4.

→ You must manually transition to the new endpoint and models, on /fine-tuning.

Old Embedding Models Retirement

All models except for text-embedding-ada-002 will be shutdown on Jan 4.

→ You must manually transition to the new model text-embedding-ada-002.

No Deprecations for Chat Models

All GPT-4 and gpt-3.5-turbo models remain active. The earliest deprecation is happening on June 13th for gpt-4-0314, gpt-4-0613, gpt-3.5-turbo-0613, gpt-3.5-turbo-0301.

Azure OpenAI will deprecate the -0314 and -0301 models on July 5th. (Azure shutdown for -0613 models is to be announced.)

→ It is recommended to start shifting your gpt-4 and gpt-3.5-turbo workloads to the newer, cheaper -1106 models.

We've put all of these updates in a simple-to-understand OpenAI Model Map. (Click on the image to expand)

For Chat Models


For Text models & /fine-tunes


Assistants API beta v1 retirement

Access to the v1 beta will be discontinued on December 18, 2024. gpt-4

→ It is recommended to start shifting your OpenAI-Beta: assistants=v1 to OpenAI-Beta: assistants=v2

Refer to OpenAI's Deprecations guide for more info.