In a significant advancement for developers and tech enthusiasts, a new version of the AI model, GPT-3.5 Turbo (designated as ‘gpt-3.5-turbo-0125’), was launched last week, featuring substantial cost reductions and several enhancements. The update has halved the input costs to $0.0005 per 1,000 tokens, while output costs have seen a 25% reduction, now priced at $0.0015 per 1,000 tokens. This latest iteration also brings minor improvements, including a crucial bug fix related to UTF-8 encoding during function calls. Starting February 15th, the default reference for ‘gpt-3.5-turbo’ will automatically direct users to this enhanced model, ensuring broader access to its upgraded capabilities.
In addition to the GPT-3.5 Turbo update, the company has been busy with other notable releases aimed at enriching the developer experience. Highlights include the updated GPT-4 Turbo preview model, known as ‘gpt-4-0125-preview’, which offers more comprehensive code generation tasks alongside other minor improvements and bug fixes. The introduction of two new embedding models marks another leap forward, featuring the ‘text-embedding-3-small’ for high efficiency and the ‘text-embedding-3-large’ for superior performance.
Developers concerned with content safety will find the updated text moderation model, ‘text-moderation-007’, particularly useful. It’s designed to assist in identifying potentially harmful text within applications more effectively. Furthermore, new management tools have been introduced to oversee and understand API usage better. These tools include the ability to assign specific permissions to API keys directly from the API keys page and the option to view detailed usage statistics at the API key level within the usage dashboard.
These updates underscore the company’s commitment to providing cutting-edge tools and resources that empower developers to create, innovate, and maintain high standards of quality and safety in their applications.