OpenAI has released new versions of its text-generating models, GPT-3.5-turbo and GPT-4, with a new capability called function calling. This allows developers to describe programming functions to the models and have them create code to execute those functions. Function calling can help create chatbots that answer questions, convert natural language into database queries, and extract structured data from text. OpenAI is also introducing a flavor of GPT-3.5-turbo with a greatly expanded context window, allowing it to consider more text before generating additional text. The new version offers four times the context length of the vanilla version at twice the price. OpenAI is reducing pricing for the original GPT-3.5-turbo by 25%, and for text-embedding-ada-002 by 75%. The reductions were made possible by increased efficiency in OpenAI’s systems. OpenAI has signaled that incremental updates to existing models, not massive new from-scratch models, are its focus. CEO Sam Altman has reaffirmed that OpenAI has not begun training the successor to GPT-4.
