OpenAI introduced function calling, referred to as “tool use,” for its GPT-4 and GPT-3.5-turbo models with an API update on June 13, 2023. The launch of OpenAI’s Assistants API on November 6, 2023, further boosted this capability with native support for function calling and persistent threads. The Responses API, announced on March 11, 2025, refined function calling into a more organized workflow. To architect effective sequences, developers should define functions as JSON Schema objects, noting a hard limit on the number of functions per API call remains unconfirmed as of July 2026. The tool_choice parameter can guide the model to use a specific function. Parallel function calling, introduced in the gpt-4-1106-preview released in November 2023, allows multiple functions to execute at once within a single model response. OpenAI’s GPT-4o, released on May 13, 2024, supports function calling with lower latency compared to earlier models.