OpenAI launches new embedding models, GPT Turbo updates

OpenAI introduced a series of updates to provide developers with more powerful tools, cost-effective solutions, and enhanced control over their API usage.

The latest models include two new embedding models, an updated GPT-4 Turbo preview model, an updated GPT-3.5 Turbo model, and an updated text moderation model. 

Data sent to the OpenAI API will not be used to train or improve OpenAI models by default.

Among the new embedding models, OpenAI unveiled text-embedding-3-small, a highly efficient and smaller model, and text-embedding-3-large, a larger and more powerful model.

The text-embedding-3-small model showcases improved performance compared to its predecessor, with a 5X reduction in pricing. 

The text-embedding-3-large model outperforms previous models, allowing developers to trade off performance and cost by shortening embeddings.

OpenAI also revealed updated GPT-3.5 Turbo and GPT-4 Turbo preview models, with reduced prices for the former. 

The GPT-4 Turbo preview model, gpt-4-0125-preview, addresses issues related to code generation and includes fixes for non-English UTF-8 generations. 

OpenAI plans to launch GPT-4 Turbo with a vision of general availability in the coming months.

In moderation, OpenAI releases text-moderation-007, its most robust moderation model to date. This model is part of OpenAI's ongoing safety efforts to identify potentially harmful text.

The update also introduces new ways for developers to understand API usage and manage API keys. 

Developers can now assign permissions to API keys and access metrics on an API key level through the usage dashboard and export function.

OpenAI's continuous commitment to improvement is evident, with plans for further enhancements in the coming months to empower developers in larger organizations to view API usage and manage API keys more effectively.