OpenAI launches new generation of embedding models and other API updates

OpenAI, the artificial intelligence research company, announced on Thursday a new generation of embedding models, which can convert text into a numerical form that can be used for various machine learning tasks. The company also introduced new versions of its GPT-4 Turbo and moderation models, new API usage management tools, and lower pricing on its GPT-3.5 Turbo model.

Embeddings are sequences of numbers that represent 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.

OpenAI said that its new embedding models, text-embedding-3-small and text-embedding-3-large, offer stronger performance and reduced price compared to its previous generation model, text-embedding-ada-002. The new models can create embeddings with up to 3072 dimensions, which can capture more semantic information and improve the accuracy of downstream tasks.

According to the company, the new models have increased the average score on a commonly used benchmark for multi-language retrieval (MIRACL) from 31.4% to 54.9%, while the average score on a commonly used benchmark for English tasks (MTEB) has increased from 61.0% to 64.6%. The pricing for text-embedding-3-small has also been reduced by 5X compared to text-embedding-ada-002, making it more affordable for developers to use.

The company also updated its GPT-4 Turbo and GPT-3.5 Turbo models, which are large multimodal models that can understand and generate natural language or code. The new versions of the models come with improved instruction following, JSON mode, more reproducible outputs, and parallel function calling. The company also launched a new 16k context version of GPT-3.5 Turbo, which can process longer inputs and outputs than the standard 4k version.

Additionally, the company updated its text moderation model, which can detect whether text may be sensitive or unsafe. The new version of the model can handle more languages and domains, and can also provide explanations for its predictions.

The company also introduced new ways for developers to manage API keys and understand API usage. Developers can now create multiple API keys with different permissions and scopes, and monitor their usage and billing details on the OpenAI Dashboard. The company also said that it will soon lower the pricing on its GPT-3.5 Turbo model by 25%, making it more accessible for developers to build applications with it.

OpenAI said that these updates are part of its continuous efforts to improve its models and services, and to make them more useful and affordable for developers and customers. The company also invited developers to contribute evaluations to help it improve the model for different use cases. The company said that it will continue to release new models, features, and tools in the future.