Steer Reasoning Performance with Effort Levels in Opus 5
Use the 'effort' parameter to control the reasoning depth of Claude Opus 5.
Control the depth of Claude Opus 5's internal reasoning by setting the 'effort' parameter to levels like 'xhigh' or 'max' for capability-critical tasks.
The Scenario
You are using Claude Opus 5 for a complex software architecture review or a high-stakes data analysis where accuracy is more important than speed. You need the model to use its full reasoning capabilities.
Before & after
Users typically had to rely on prompt engineering (e.g., 'think step by step') to encourage deeper reasoning, which is inconsistent and takes 5–10 minutes of prompt testing.
By choosing an effort level from 'low' to 'max', you can now explicitly tell Claude Opus 5 how much 'thinking' time to spend on a problem, with the 'max' setting taking 1-2 minutes for the highest accuracy.
The Prompt
// Adjust the effort level in your API call to control reasoning depth
{
"model": "claude-opus-5",
"effort": "max",
"thinking": {"type": "enabled"},
"messages": [{"role": "user", "content": "[INSERT_COMPLEX_PROBLEM_HERE]"}]
}Claude Opus 5 uses 'thinking' by default. 'Effort' is the primary control for steering this behavior. It supports a ladder: low, medium, high, xhigh, and max. For capability-critical work, 'max' should be used.
Source
Claude Platform release notes - Claude Platform Docs"Effort is the primary control for steering Claude Opus 5: the model supports the full ladder (low, medium, high, xhigh, max)."