
OpenAI has unveiled a series of enhancements aimed at developers using its AI solutions, including new embedding models, a reduced cost for GPT-3.5 Turbo, an updated version of GPT-4 Turbo, and enhanced content moderation features.
The AI research company based in San Francisco has introduced its latest text-embedding-3-small and text-embedding-3-large models, which exhibit superior performance compared to their predecessors. Notably, the text-embedding-3-large model shows significant improvements, scoring 54.9 percent on the MIRACL benchmark and 64.6 percent on the MTEB benchmark, a marked increase from the 31.4 percent and 61 percent scores of the older text-embedding-ada-002 model.
OpenAI has also significantly reduced the price for its text-embedding-3-small model by 5x per 1,000 tokens, down from $0.0001 to $0.00002. The company highlights that developers can now optimize costs by shortening embeddings, with minimal effect on accuracy.
In the upcoming week, OpenAI intends to launch an enhanced version of GPT-3.5 Turbo and slash its pricing by half for input tokens and by 25 percent for output tokens. This adjustment is part of a series of price reductions aimed at increasing the model’s adoption.
Moreover, the GPT-4 Turbo preview has been upgraded to version gpt-4-0125-preview, with over 70 percent of requests already utilizing the new model since its introduction. The upgrade focuses on improving code generation and other functionalities.
For developers focused on creating safe AI applications, OpenAI has introduced its most sophisticated content moderation model to date, text-moderation-007, which outperforms earlier versions in identifying potentially harmful content.
Additionally, OpenAI has enhanced API key management and usage metrics visibility, allowing developers to assign specific permissions to keys and monitor usage at an individual key level, facilitating better tracking of products or projects.
Source: AINews