result = Runner.run_streamed(
starting_agent=manager_agent,
input=“User prompt goes here",
run_config=RunConfig(model=“gpt-4.1-nano"),
)
I’m following the manager (agents as tools) pattern where I have a manager agent that can use a few other agents as tools. I do not set model on any of the agents and only set the model in the run_config. Should I only be seeing gpt-4.1-nano in my usage? I believe the documentation says the model should be used throughout the entire run?
I actually see both gpt-4_1-nano-2025-04-14 and gpt-4o-2024-08-06 in my usage stats.
Hi!
I understand that you expect defining a run_config will globally set all agents to always use the defined configuration.
run_config is defined as:
Global settings for the entire agent run
The global scope applies to a single agent, not a whole set of agents.
If the model for an agent is not defined, 4o will be used as the default.
As a solution for now, you can define the model explicitly for each agent.
I think this can be more clear and will re-categorize this topic to Documentation. It’s possible that this is a bug. Otherwise it would be helpful to better define global in this context of agents as tools, IMO.