# Ultrathink

> A word you add to one prompt so Claude thinks harder on that turn only.

**Field:** Artificial Intelligence  
**Category:** Claude Code  
**Also known as:** ultrathink keyword, Claude Code think keyword

## Think of it like

It is like asking a friend to go slow. This one puzzle is hard.

## Definition

Ultrathink is a word you put in a Claude Code prompt to ask for deeper thinking on that one turn. Claude Code spots the word and adds a note asking the model to reason harder. In the terminal it shows in rainbow colours when it can also raise that turn to the xhigh effort level. Your normal effort setting comes back on the next turn. Other phrases like think hard are passed on as plain text.

## Example

Leo's tests fail in a strange way. He writes ultrathink: why does this test pass alone but fail in the suite? Claude spends longer on that one answer.

## Example code

```console
> ultrathink: plan the move from REST to gRPC
```

One word raises the thinking for this turn, then the setting goes back.

## Why it matters

Hard turns get deep thought, and easy turns stay fast and cheap. You never have to change the session setting.

## Common mistake

Typing think hard and expecting the same boost. Only the word ultrathink is treated as a keyword.

## Three principles

- **One turn only.** The session effort stays as it was.
- **Rainbow means it works.** It can raise the turn to xhigh.
- **Only this word.** Other think phrases are plain text.

## Related terms

- [Claude Code Effort Level](https://learn.benibix.com/term/ai/claude-code-effort-level)
- [CLAUDE_CODE_DISABLE_ULTRATHINK](https://learn.benibix.com/term/ai/claude-code-disable-ultrathink-variable)
- [Extended Thinking](https://learn.benibix.com/term/ai/extended-thinking)
- [Test-Time Compute](https://learn.benibix.com/term/ai/test-time-compute)
- [/model Command](https://learn.benibix.com/term/ai/model-command)

## Told apart from

- [Ultrathink vs Effort Level](https://learn.benibix.com/compare/ultrathink-vs-claude-code-effort-level): People think ultrathink is one more effort level. It is not. The effort level sets how hard Claude thinks for the whole session, from low to max. Typing ultrathink in a prompt asks for deeper thinking on that one prompt only. The next prompt goes back to the effort level you set.

## Taught in

- [Claude Code Models and Effort](https://learn.benibix.com/learn/ai-claude-code/claude-code-models-and-effort): Claude Code, 8/16

![A scale of five effort levels from low to max. The low end is fast and cheap. The high end is deep, slow and costly.](https://learn.benibix.com/diagrams/cc-effort-level.svg)

---

Source: https://learn.benibix.com/term/ai/ultrathink
Author: Behnia Zarei (a developer in Vienna, Austria, who builds and teaches data, AI and business intelligence at BENIBIX)
Publisher: BENIBIX Learn (https://learn.benibix.com/)
Last updated: 2026-11-19
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
