Skip to content
Cost & Performance Simple

Effort Levels Comparison

Control reasoning depth with --effort: low for speed, max for thoroughness

Command

$ "color:#7C5CFC">claude -p "What is 2+2?" "color:#d97757">--effort low "color:#d97757">--output-format json

Response

| Level  | Behavior               | Best for                    |
|--------|------------------------|-----------------------------||
| low    | Minimal reasoning      | Lookups, yes/no, classification |
| medium | Balanced reasoning     | Formatting, summaries       |
| high   | Full reasoning (default)| Dev tasks, code gen         |
| max    | Maximum depth          | Architecture, subtle bugs   |

Parsing Code

059669">">// Set via flag, env 059669">">var, or CLAUDE.md:
claude -p "Task059669">" --effort medium

059669">">// Environment variable:
059669">">export CLAUDE_EFFORT=medium

059669">">// CLAUDE.md:
// effort: medium

Gotchas

! low = fastest and cheapest, max = most thorough but most expensive
! Can be set permanently in CLAUDE.md with 'effort: medium' for project-wide defaults

Related Recipes