This article is only relevant if you started using the API before January 20, 2022.
Summary
The Instruct series, previously in beta, are now our recommended GPT-3 models, and we're updating our engine names as part of this change.
New GPT-3 models
In 2021, we released a set of beta GPT-3 models called the Instruct series. The Instruct models, unlike our original base GPT-3 models, are optimized to follow your instructions. This means they're better at producing accurate completions for your prompts, as you can see below.
Base model example | Instruct model example |
Prompt Explain the moon landing to a 6 year old in a few sentences.
Completion Explain the moon landing to a 6 year old in a few sentences.
Explain the moon landing to a 6 year old in a few sentences.
Explain the moon landing to a 6 year old in a few sentences. | Prompt Explain the moon landing to a 6 year old in a few sentences.
Completion In 1969, the United States sent a spacecraft called Apollo 11 to land on the moon. Astronauts named Neil Armstrong and Edwin "Buzz" Aldrin were the first people to step out of the spacecraft and walk on the moon. |
After overwhelmingly positive feedback, we're making these models our primary GPT-3 models and removing the "Instruct" descriptor. Going forward, we recommend using them for all natural language tasks. For most use cases, we expect that you can use them as a drop-in replacement for the base models with no changes to your prompt.
The base GPT-3 models are still available, but we don't recommend using them unless you're fine-tuning a custom model or using our search, answers, or classification endpoints.
New engine names
As part of this change, we've also updated most of our engine names to be clearer and support adding more models in the future. API requests that specify the old engine names will continue working as expected, but we recommend that you start using the new names going forward.
Type | Old engine name | New engine name |
Latest Instruct GPT-3 models | davinci-instruct-beta-v3 curie-instruct-beta-v2 babbage-instruct-beta ada-instruct-beta | text-davinci-001 text-curie-001 text-babbage-001 text-ada-001 |
Older Instruct GPT-3 models | davinci-instruct-beta curie-instruct-beta | No change |
Base GPT-3 models | davinci curie babbage ada | No change |
Codex models | davinci-codex cushman-codex | code-davinci-001 code-cushman-001 |
Similarity embeddings models | ada-similarity babbage-similarity curie-similarity davinci-similarity | text-similarity-ada-001 text-similarity-babbage-001 text-similarity-curie-001 text-similarity-davinci-001
|
Text search embeddings models | ada-search-document ada-search-query babbage-search-document babbage-search-query curie-search-document curie-search-query | text-search-ada-doc-001 text-search-ada-query-001 text-search-babbage-doc-001 text-search-babbage-query-001 text-search-curie-doc-001 text-search-curie-query-001 |
Code search embeddings models | ada-code-search-code ada-code-search-text babbage-code-search-code babbage-code-search-text
| code-search-ada-code-001 code-search-ada-text-001 code-search-babbage-code-001 code-search-babbage-text-001 |
Content filter | content-filter-alpha | No change |
If you have any questions about these updates, please log into your account and start a new conversation using the on-site chat tool. Alternatively, ask a question on our community forum.