OpenAI has recently announced updates in their API offerings, introducing new embedding models, an updated GPT-4 Turbo, moderation models, new API management tools, and reduced pricing for GPT-3.5 Turbo.
Key Takeaways:
- New Embedding Models: OpenAI introduced two new embedding models – text-embedding-3-small and text-embedding-3-large. The small model is more efficient and cheaper, improving the multi-language retrieval (MIRACL) score from 31.4% to 44.0%, and the English tasks (MTEB) score from 61.0% to 62.3%. The larger model provides up to 3072 dimensions, significantly enhancing performance on benchmarks.
- Reduced Pricing: The new text-embedding-3-small model’s pricing has been reduced by 5X compared to its predecessor, while the text-embedding-3-large will cost $0.00013 per 1k tokens.
- GPT-3.5 Turbo Updates and Price Reduction: A new GPT-3.5 Turbo model is being introduced, with a 50% reduction in input prices and a 25% reduction in output prices. It also includes improvements for higher accuracy and bug fixes for non-English language function calls.
- Updated GPT-4 Turbo Preview: The new GPT-4 Turbo model will offer updated knowledge cutoff, larger context windows, and reduced prices. It’s designed to complete tasks like code generation more thoroughly and address issues of task completion.
- New Moderation Model: OpenAI is releasing an updated moderation model, text-moderation-007, which is more robust and efficient in identifying potentially harmful text.
- Future Plans: OpenAI plans to launch GPT-4 Turbo with vision in the coming months and improve API usage management and key controls, especially for larger organizations.
- Enhanced API Management: Developers now have improved tools to manage API keys and understand usage, including assigning permissions to keys and tracking usage metrics at an API key level.
OpenAI dashboard usage:
