OpenAI Introduces New API Upgrades
OpenAI makes some notable improvements to the embeddings and text completion API's!
Created on January 26|Last edited on January 26
Comment
OpenAI has announced a series of significant updates and releases, including new embedding models, an improved GPT-4 Turbo model, lower pricing on GPT-3.5 Turbo, and enhanced tools for API usage management.
Enhanced Embedding Models and Reduced Costs
OpenAI has unveiled two new embedding models: text-embedding-3-small and text-embedding-3-large. These models are designed to understand content and context more effectively, aiding in tasks like clustering or retrieval.
text-embedding-3-small: This model offers improved performance over its predecessor, text-embedding-ada-002, and comes at a lower price. It shows notable improvements in benchmarks like MIRACL and MTEB, signifying better multi-language retrieval and English task handling.
text-embedding-3-large: This larger model creates up to 3072-dimensional embeddings, providing superior performance on the same benchmarks. It is priced competitively at $0.00013 per 1k tokens.
Both models allow developers to adjust embedding sizes, balancing performance with computational and storage costs.
Updates in GPT Models
GPT-3.5 Turbo: The upcoming gpt-3.5-turbo-0125 model features a significant price reduction and improvements in accuracy and format response. The price reduction aims to facilitate scalability for customers.
GPT-4 Turbo: The updated gpt-4-0125-preview model addresses previous limitations by improving task completion and reducing instances of incomplete responses. The model has also been enhanced to better handle non-English UTF-8 generations.
New Moderation Model
This robust moderation model, accessible via the free Moderation API, helps developers identify potentially harmful text more effectively.
API Usage and Key Management Tools
OpenAI introduced new tools for API key management and usage understanding. Developers can now assign specific permissions to API keys and track usage metrics at the API key level. These features are particularly beneficial for managing access and understanding usage patterns in larger organizations.
Future Plans and Availability
OpenAI plans to launch GPT-4 Turbo with vision in general availability in the coming months, indicating ongoing efforts to broaden the capabilities of their AI models.
Developers can access detailed information and guides on using these new models and tools on OpenAI’s website.
The announcement: https://openai.com/blog/new-embedding-models-and-api-updates
Add a comment
Tags: ML News
Iterate on AI agents and models faster. Try Weights & Biases today.