How GPT Decides to Call Tools

~22 min · tool-decision, tool_choice

Level 0Tokenizer

0 XP0/54 lessons0/10 achievements

0/120 XP to next level120 XP to go0% complete

The tool_choice parameter controls whether and how the model uses tools.

Value	Behavior
`"auto"` (default)	Model decides whether to call zero, one, or multiple tools
`"required"`	Model must call at least one tool
`"none"`	Model cannot call any tools (text-only response)
`{"type":"function","name":"get_weather"}`	Force a specific function call
`{"type":"allowed_tools","mode":"auto","tools":[...]}`	Restrict to a subset of available tools

With "auto", the model examines the user's query and available tool descriptions to decide. Good tool descriptions are critical — they're the model's "manual" for knowing when and how to use each tool.

Code

tool_choice='auto' vs 'required' vs specific·python

# Force a specific function call
response = client.responses.create(
    model="gpt-5.4",
    input="What's the weather?",
    tools=tools,
    tool_choice={"type": "function", "name": "get_weather"},
)

# Require at least one tool call
response = client.responses.create(
    model="gpt-5.4",
    input="Look up current events.",
    tools=tools,
    tool_choice="required",
)

External links

tool_choice

Exercise

For one tool, run the same prompt three times: tool_choice='auto', 'required', and the specific name. Note which inputs make the model skip the tool under 'auto' and which force it under 'required'.