OpenAI, a well-known artificial intelligence research center, has unveiled an update to its language model API, including the highly anticipated GPT-4 model and gpt-3.5-turbo-16k. These models aim to expand the context window, which functions as a short-term memory, allowing developers to process more text simultaneously. OpenAI is also developing a 32,000-token variant of GPT-4, which remains unavailable to the public. Additionally, the company has introduced new features to its API, such as function call functionality to create chatbots capable of calling external tools, converting natural language into API calls, or conducting database queries.
OpenAI has also enhanced the controllability of its models, allowing users to influence the behavior of the language model. With the new gpt-3.5-turbo-0613 model, users can influence the behavior of the language model more efficiently. These updates have also significantly reduced the costs associated with using these models. In particular, the cost of the gpt-3.5-turbo model has been reduced by 25%, making it more affordable and accessible to developers.
Moreover, the cost of the text-embedding-ada-002 model, which enables computers to comprehend words and concepts, has also been substantially reduced by 75%. OpenAI has announced that it will retire some of its earlier model variants starting September 13. However, until that date, developers may still use these models.
With these new updates, OpenAI aims to reduce developers’ costs and provide better text generation and modeling efficiency to users. These updates offer various features such as function call functionality, enhanced controllability, and more affordable pricing, making the OpenAI language model API a more accessible tool for developers.
