AI BriefingNew Embedding Models and API Updates
Key point
OpenAI announced new embedding models with improved performance and lower pricing, along with a GPT-3.5 Turbo update.
Details
OpenAI announced new Embedding models and a GPT-3.5 Turbo update that significantly improve performance and lower costs. Data sent via the API is not used to train models by default.
The new embedding models consist of two models:
- text-embedding-3-small: Performance is improved compared to the existing
text-embedding-ada-002, and pricing is 5x cheaper ($0.0001 $\rightarrow$ $0.00002/1k tokens). - text-embedding-3-large: A next-generation model supporting up to 3,072 dimensions, overwhelming the existing model with a score of 54.9% on the MIRACL benchmark.
In particular, these new models support a Dimensions parameter feature. Through the dimensions parameter, developers can reduce the dimensions of embeddings, allowing them to efficiently manage cost and storage space at the expense of some accuracy.
Also, pricing for the GPT-3.5 Turbo model (gpt-3.5-turbo-0125) will be reduced. Input token pricing is lowered by 50% ($0.0005/1k tokens) and output token pricing by 25% ($0.0015/1k tokens), reducing operating costs for large-scale services.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.