New embedding models and API updates

OpenAI

We are releasing new models, reducing prices for GPT‑3.5 Turbo, and introducing new ways for developers to manage API keys and understand API usage. The new models include:

• Two new embedding models

• An updated GPT‑4 Turbo preview model 

• An updated GPT‑3.5 Turbo model

• An updated text moderation model

By default, data sent to the OpenAI API will not be used to train or improve OpenAI models.

New embedding models with lower pricing

We are introducing two new embedding models: a smaller and highly efficient text-embedding-3-small model, and a larger and more powerful text-embedding-3-large model.

An embedding⁠ is a sequence of numbers that represents the concepts within content such as natural language or code. Embeddings make it easy for machine learning models and other algorithms to understand the relationships between content and to perform tasks like clustering or retrieval. They power applications like knowledge retrieval in both ChatGPT and the Assistants API, and many retrieval augmented generation (RAG) developer tools.

A new small text embedding model

text-embedding-3-small is our new highly efficient embedding model and provides a significant upgrade over its predecessor, the text-embedding-ada-002 model released in December 2022⁠. 

Stronger performance. Comparing text-embedding-ada-002 to text-embedding-3-small, the average score on a commonly used benchmark for multi-language retrieval (MIRACL⁠(opens in a new window)) has increased from 31.4% to 44.0%, while the average score on a commonly used benchmark for English tasks (MTEB⁠(opens in a new window)) has increased from 61.0% to 62.3%.

Reduced price. text-embedding-3-small is also substantially more efficient than our previous generation text-embedding-ada-002 model. Pricing for text-embedding-3-small has therefore been reduced by 5X compared to text-embedding-ada-002, from a price per 1k tokens of $0.0001 to $0.00002.

A new large text embedding model: text-embedding-3-large

Native support for shortening embeddings

1. A