Introduction
OpenAI has recently announced a series of exciting updates!
New Embedding Models
• Small Text Embedding Model (text-embedding-3-small)
- Upgraded Performance: Significant improvement over its predecessor, text-embedding-ada-002. Performance on MIRACL increased from 31.4% to 44.0%, and on MTEB from 61.0% to 62.3%.
- Reduced Pricing: 5X cheaper than text-embedding-ada-002, priced at $0.00002 per 1k tokens.
- Continued Support for text-embedding-ada-002: For those who prefer the previous model.
Large Text Embedding Model (text-embedding-3-large)
- Enhanced Capabilities: Capable of creating embeddings up to 3072 dimensions.
- Superior Performance: Outperforms text-embedding-ada-002 on MIRACL (54.9%) and MTEB (64.6%).
- Cost-effective Pricing: Available at $0.00013 per 1k tokens.
Flexibility in Embeddings Usage
The new models allow developers to adjust embedding sizes for a balance between performance and cost.
👁 Updated Models and Lower Pricing
GPT-3.5 Turbo Model (gpt-3.5-turbo-0125)
- Price Reduction: Input prices cut by 50% to $0.0005 per 1K tokens, and output prices by 25% to $0.0015 per 1K tokens.
- Enhanced Accuracy: Improved response format accuracy and a fix for a non-English language text encoding issue.
GPT-4 Turbo Preview Model (gpt-4-0125-preview)
- Advanced Capabilities: Better at completing tasks like code generation and fixing previous model limitations.
- Continuous Upgrades: The introduction of a new model alias for automatic updates to the latest GPT-4 Turbo model.
New Developer Tools
• API Key Management: Developers can now assign specific permissions to API keys for more controlled access.
• API Usage Insights: Enhanced tracking and reporting features for API usage, enabling better management and optimization.
🛡 Enhanced Safety with Updated Moderation Model
The new text-moderation-007 model, with improved capabilities to identify harmful text.
Source rozetka.pl
Top comments (0)