- OpenAI introduces function calling capabilities, updated versions of gpt-4 and gpt-3.5-turbo, and a 16k context version of gpt-3.5-turbo.
- The pricing for the API has been reduced, with a 75% cut for the embeddings model and a 25% cut for gpt-3.5-turbo.
- The gpt-3.5-turbo-0301 and gpt-4-0314 models will be deprecated, with the deprecation timeline announced.
OpenAI has announced a series of updates aimed at enhancing the performance and usability of its models. The new improvements include function calling capabilities in the Chat Completions API, enhanced and more steerable versions of gpt-4 and gpt-3.5-turbo, a 16k context version of gpt-3.5-turbo, and significant cost reductions on certain models. The company is also declaring the deprecation timeline for the gpt-3.5-turbo-0301 and gpt-4-0314 models.
The function calling capability has been introduced in the gpt-4-0613 and gpt-3.5-turbo-0613 models. It allows developers to describe functions to these models, which can then intelligently output a JSON object containing arguments to call those functions. This improves the connectivity of the GPT capabilities with external tools and APIs. By responding with JSON that adheres to the function signature, the models can more reliably provide structured data.
With the function calling capability, developers can create chatbots that answer questions by calling external tools, convert natural language into API calls or database queries, and extract structured data from text. The capability is enabled by new API parameters that allow developers to describe functions to the model via JSON Schema.
The gpt-4-0613 and gpt-4-32k-0613 models include the same function calling capability, with the latter also offering extended context length for better comprehension of larger texts. The gpt-3.5-turbo-0613 model includes the same function calling as GPT-4 and more reliable steerability via the system message.
The initial versions of gpt-4 and gpt-3.5-turbo announced in March will begin the deprecation process. Developers can continue using the older models until September 13th, after which requests specifying those model names will fail.
The cost of the embeddings model, text-embedding-ada-002, has been reduced by 75%. The input token cost for the gpt-3.5-turbo model has been reduced by 25%.