29-effort-reasoning-control

Control Claude Code reasoning effort by configuring low, medium, high, or max effort levels at the session or frontmatter skill level. This optimizes the balance between task quality, latency, and token cost instead of applying maximum reasoning to every task.

Claude Code reasoning effort levels include low, medium, high, and max configurations. Understanding effort precedence and supported model capabilities allows you to apply the appropriate reasoning depth for your specific workflow requirements.

Vertex AI deployments support adaptive reasoning configuration for Claude models, allowing you to manage effort availability and troubleshoot missing effort options. You must understand specific Vertex AI model behavior to properly configure reasoning levels.

Missing reasoning effort options in Claude Code often stem from unsupported model capabilities or incorrect effort precedence settings. Troubleshoot by verifying your current model's supported effort levels and checking your session or frontmatter configuration.

Use ultrathink for complex workloads requiring maximum reasoning depth, while applying lower effort settings to reduce unnecessary reasoning costs for simpler tasks. Configuring effort controls at the skill level preserves quality where needed while optimizing latency.