Which model first introduced function calling?

• A. GPT-3.5
• B. GPT-4
• C. BERT
• D. T5

Answer: (B)

The key idea here is a model’s ability to call external tools or APIs during a conversation. Function calling lets the model decide which tool to use and output a structured call (with a function name and arguments) instead of just plain text. This lets the model perform actions like querying data, performing calculations, or pulling information from services, and then incorporate the results back into the conversation.

GPT-4 is the model that first introduced this capability. It brought a built-in mechanism for defining available functions and for the model to invoke them when it’s the right tool for the task. This marks a step beyond purely generating text, enabling reliable tool use within chat.

GPT-3.5 did not include this function-calling capability, so it couldn’t emit structured function calls to external tools. BERT and T5 are different kinds of models—BERT is mainly an encoder for understanding tasks, and T5 is a text-to-text model—but neither was introduced with a native framework for invoking external functions in a chat setting.