Subscribe to our daily and weekly newsletters to stay updated on industry-leading AI developments.
OpenAI, the artificial intelligence research company, announced new embedding models on Thursday. These models convert text into numerical forms for various machine learning tasks. OpenAI also introduced new versions of its GPT-4 Turbo and moderation models, new API management tools, and lowered the prices for its GPT-3.5 Turbo model.
Embeddings are numeric sequences representing concepts in content like natural language or code. They help machine learning models understand relationships within content, aiding in tasks like clustering or retrieval. These embeddings are used in applications such as knowledge retrieval in ChatGPT and the Assistants API, as well as many retrieval augmented generation (RAG) developer tools.
OpenAI’s new embedding models, text-embedding-3-small and text-embedding-3-large, outperform the previous generation model, text-embedding-ada-002, and are more cost-effective. These models create embeddings with up to 3072 dimensions, capturing more semantic information and improving task accuracy.
The new models have significantly improved benchmark scores. For multi-language retrieval (MIRACL), the average score increased from 31.4% to 54.9%, and for English tasks (MTEB), from 61.0% to 64.6%. The pricing for text-embedding-3-small is now five times lower than for text-embedding-ada-002, making it more affordable for developers.
Additionally, OpenAI updated its GPT-4 Turbo and GPT-3.5 Turbo models. These large multimodal models can understand and generate natural language or code. The new versions feature improved instruction following, JSON mode, more consistent outputs, and parallel function calling. A new 16k context version of GPT-3.5 Turbo can process longer inputs and outputs compared to the standard 4k version.
OpenAI also enhanced its text moderation model, which now detects sensitive or unsafe text more effectively. The updated model supports more languages and domains and can provide explanations for its predictions.
To help developers manage API keys and usage, OpenAI introduced new management tools. Developers can create multiple API keys with different permissions, monitor usage, and view billing details on the OpenAI Dashboard. Additionally, OpenAI will soon reduce the price of GPT-3.5 Turbo by 25%, making it more accessible for developers.
These updates are part of OpenAI’s ongoing efforts to improve its models and services, making them more useful and affordable. The company invites developers to contribute evaluations to enhance the models for various use cases. OpenAI plans to continue releasing new models, features, and tools in the future.