Configuration options for the model's output.
Supports the following keys:
effort: Controls how many tokens Claude uses when responding.
One of "max", "xhigh", "high", "medium", or "low".format: Structured output format configuration (typically set via
with_structured_output).task_budget: Advisory token budget for an agentic loop (beta).
E.g., #!python {"type": "tokens", "total": 128_000}.Example:
.. code-block:: python
ChatAnthropic(
model="claude-opus-4-7",
output_config={
"effort": "xhigh",
"task_budget": {"type": "tokens", "total": 128_000},
},
)
See Anthropic docs on extended output.