OpenAI has announced a series of exciting updates, including the release of new models, reduced pricing for the GPT-3.5 Turbo, and innovative ways for developers to handle API keys and monitor API usage.
The fresh additions to OpenAI’s lineup feature two new embedding models: an enhanced GPT-4 Turbo preview model and an updated GPT-3.5 Turbo model, along with a refined text moderation model.
Affordable and Efficient New Embedding Models
OpenAI introduces two groundbreaking embedding models:
- Text-Embedding-3-Small: A compact yet highly efficient model, offering superior performance at a significantly reduced cost, is now priced 5X lower than its predecessor.
- Text-Embedding-3-Large: A more robust model creates larger embeddings, leading the pack in performance.
These embedding models play a crucial role in understanding and correlating content across diverse formats, like natural language and code. They are integral in applications such as knowledge retrieval and support a range of developer tools.
The new small text embedding model, text-embedding-3-small, surpasses its previous version in performance benchmarks while being substantially more affordable. The large text embedding model, text-embedding-3-large, sets a new standard for embedding size and performance, excelling in multi-language retrieval and English tasks.
Developers now have the option to adjust the size of embeddings without compromising their effectiveness. This flexibility allows for cost-effective and efficient usage of embeddings in various applications.
Upgraded GPT Models and Reduced Pricing
- The new GPT-3.5 Turbo Model will launch soon with a 50% price reduction for inputs and 25% for outputs, featuring improved accuracy and bug fixes.
- GPT-4 Turbo Preview Update: More thorough task completion and bug fixes, with a new alias for automatic upgrades to the latest version.
- The upcoming GPT-4 Turbo with Vision is set to be generally available shortly.
The updated text-moderation-007 model, now more robust than ever, is part of OpenAI’s commitment to safe AI development. Developers can access this model for free to detect harmful text.
Developers will benefit from new platform enhancements, allowing for more granular control over API keys and detailed usage tracking. These features are especially beneficial for larger organizations and complex projects.
In conclusion, OpenAI’s recent updates mark a significant advancement in the field of AI, offering developers more powerful, efficient, and cost-effective tools. The introduction of new embedding models, namely text-embedding-3-small and text-embedding-3-large, represents a leap in performance and affordability, catering to a diverse range of AI applications.