Extended Thinking and Effort Levels

What You'll Learn

Not every task needs the same level of reasoning. In this lesson, you'll learn when to engage extended thinking for better results on complex problems, and when to keep things fast for simple tasks.

What Extended Thinking Actually Does

When thinking is enabled, Claude doesn't just jump to an answer. It works through the problem internally first, reasoning step by step before responding. You'll see a "Thinking..." phase in the output where this is happening.

Modern Claude models (Opus 4.6, Sonnet 4.6) use adaptive thinking. Instead of a fixed thinking budget, Claude evaluates the complexity of your request and decides how much reasoning it needs. A simple rename gets almost no thinking. A complex refactor gets deep, multi-step reasoning. This happens automatically.

Adaptive thinking also enables interleaved thinking, meaning Claude can think between tool calls during multi-step tasks. When Claude reads a file, it can reason about what it found before deciding what to do next. This makes a real difference in agentic workflows where Claude is exploring your codebase and making decisions along the way.

Toggling Extended Thinking

Inside Claude Code, press:

Shortcut	What It Does
Alt+T / Option+T	Toggle extended thinking on/off

When you toggle thinking on, Claude will reason through problems before acting. When it's off, Claude responds directly without the thinking phase, which is faster but less thorough for complex tasks.

Effort Levels

Beyond the toggle, you can control how much reasoning Claude applies. Effort levels guide how deeply Claude thinks:

Effort	Thinking Behaviour
max	Always thinks with no constraints. Opus only.
high (default)	Always thinks. Deep reasoning on complex tasks.
medium	Moderate thinking. May skip it for very simple queries.
low	Minimal thinking. Skips it where speed matters most.

At high effort, Claude will almost always engage its thinking process. At low, it will skip thinking entirely for straightforward tasks and only engage it when the problem genuinely requires reasoning.

When to Use What

Higher effort (high / max) for:

• Architectural decisions ("How should I structure the data layer?")
• Complex refactors that touch multiple files
• Debugging subtle issues where the root cause isn't obvious
• Tasks that require weighing trade-offs

Lower effort (low / medium) for:

• Simple edits ("Rename this variable")
• Adding a straightforward component
• Running commands or scripts
• Tasks with obvious, single solutions

You'll develop an intuition for this as you use Claude Code more. The key insight is that effort and cost are directly related. Higher effort means more tokens, which means more cost and latency. Not every task needs deep analysis.

Key Takeaways

• Extended thinking lets Claude reason through problems before responding
• Modern models use adaptive thinking, automatically scaling reasoning to task complexity
• Interleaved thinking means Claude can reason between tool calls during multi-step workflows
• Alt+T toggles thinking on/off in Claude Code
• Effort levels (low through max) let you control how deeply Claude thinks
• Match effort to task complexity. Speed and cost matter too

Next up: One more lesson in this section. We'll take everything you've built and give it a framework that makes your instructions work harder.