GPT Rate limits, imposed by an API to control the frequency of user or client server access within a specified timeframe, are accessible under the rate limits section on the account management page for any given organization.
As GPT-4 is rolled out, it will exhibit stricter rate limits in order to meet the demand. The current rate limits can be inspected under the rate limits section on the account page. However, due to limited capacity, requests for increases in rate limits cannot be entertained at this moment. Initially, the priority lies in granting general access to GPT-4. Later, as the capacity permits, the rate limits will be gradually increased.
OpenAI applies rate limits to the number of requests made to its API. The limitations could be enforced on a per-minute basis across several parameters like requests, tokens, or in the scenario of image models, images.
Default Organization Rate Limits
Rate limits are measured in two ways: RPM (requests per minute) and TPM (tokens per minute).
| MODEL | RPM | TPM |
|---|---|---|
| CHAT | ||
| gpt-3.5-turbo | 3,500 | 90,000 |
| gpt-3.5-turbo-0301 | 3,500 | 90,000 |
| gpt-3.5-turbo-0613 | 3,500 | 90,000 |
| gpt-3.5-turbo-16k | 3,500 | 180,000 |
| gpt-3.5-turbo-16k-0613 | 3,500 | 180,000 |
| gpt-4 | 200 | 40,000 |
| gpt-4-0314 | 200 | 40,000 |
| gpt-4-0613 | 200 | 40,000 |
| TEXT | ||
| ada | 3,000 | 250,000 |
| ada-code-search-code | 3,000 | 250,000 |
| ada-code-search-text | 3,000 | 250,000 |
| ada-search-document | 3,000 | 250,000 |
| ada-search-query | 3,000 | 250,000 |
| ada-similarity | 3,000 | 250,000 |
| babbage | 3,000 | 250,000 |
| babbage-code-search-code | 3,000 | 250,000 |
| babbage-code-search-text | 3,000 | 250,000 |
| babbage-search-document | 3,000 | 250,000 |
| babbage-search-query | 3,000 | 250,000 |
| babbage-similarity | 3,000 | 250,000 |
| code-davinci-edit-001 | 20 | 150,000 |
| code-search-ada-code-001 | 3,000 | 250,000 |
| code-search-ada-text-001 | 3,000 | 250,000 |
| code-search-babbage-code-001 | 3,000 | 250,000 |
| code-search-babbage-text-001 | 3,000 | 250,000 |
| curie | 3,000 | 250,000 |
| curie-instruct-beta | 3,000 | 250,000 |
| curie-search-document | 3,000 | 250,000 |
| curie-search-query | 3,000 | 250,000 |
| curie-similarity | 3,000 | 250,000 |
| davinci | 3,000 | 250,000 |
| davinci-instruct-beta | 3,000 | 250,000 |
| davinci-search-document | 3,000 | 250,000 |
| davinci-search-query | 3,000 | 250,000 |
| davinci-similarity | 3,000 | 250,000 |
| text-ada-001 | 3,000 | 250,000 |
| text-babbage-001 | 3,000 | 250,000 |
| text-curie-001 | 3,000 | 250,000 |
| text-davinci-001 | 3,000 | 250,000 |
| text-davinci-002 | 3,000 | 250,000 |
| text-davinci-003 | 3,000 | 250,000 |
| text-davinci-edit-001 | 20 | 150,000 |
| text-embedding-ada-002 | 3,000 | 1,000,000 |
| text-search-ada-doc-001 | 3,000 | 250,000 |
| text-search-ada-query-001 | 3,000 | 250,000 |
| text-search-babbage-doc-001 | 3,000 | 250,000 |
| text-search-babbage-query-001 | 3,000 | 250,000 |
| text-search-curie-doc-001 | 3,000 | 250,000 |
| text-search-curie-query-001 | 3,000 | 250,000 |
| text-search-davinci-doc-001 | 3,000 | 250,000 |
| text-search-davinci-query-001 | 3,000 | 250,000 |
| text-similarity-ada-001 | 3,000 | 250,000 |
| text-similarity-babbage-001 | 3,000 | 250,000 |
| text-similarity-curie-001 | 3,000 | 250,000 |
| text-similarity-davinci-001 | 3,000 | 250,000 |
| MODERATION | ||
| text-moderation-latest | 1,000 | 1,250 |
| text-moderation-stable | 1,000 | 1,250 |
| IMAGE | IMG / MIN | |
| DALL·E 2 | ∞ | 50 |
| AUDIO | ||
| whisper-1 | 50 | 25,000,000 |
| OTHER | ||
| Default limits for all other models | 3,000 | 250,000 |
Read more related articles: