Match the effort level to the task

What the levels mean

The levels depend on the agent:

• Claude Code: Low, Medium, High, xHigh, Max, and Ultracode. Medium is the default.
• Codex: Low, Medium, High, Extra high, Max, and Ultra. Medium is the default.
• Grok Build: Low, Medium, High, and Extra high. High is the default.
• Antigravity: Model default, Low, Medium, and High.
• OpenCode: Model default only.

Claude Code’s documentation describes the main levels this way:

• Low suits “quick exchanges where you review each result, such as brainstorming, a first sketch, or a small change like a rename.”
• Medium, Claude’s default on Opus 5.5 and Sonnet 5.5, fits “day-to-day engineering work with a clear scope, such as implementing a new feature.”
• High suits “work where verification matters or edge cases are likely, such as fixing a bug in an existing codebase.”
• Max is for “hard problems you want Claude to work through without you, such as finding security vulnerabilities.” The same page warns it “may show diminishing returns and is prone to overthinking.”

The top levels work differently. Codex says “Ultra uses subagents to handle separate parts of a complex task in parallel,” and suggests choosing it “when you can divide the work into meaningful parts.” Claude Code’s Ultracode has Claude plan a multi-step workflow for each substantial task. Both run extra agents behind the scenes, and each one draws on your plan, so save them for big jobs that split into parts.

How to choose in Enjoy

1. Start at the default. Codex’s documentation says to “start with the default effort and increase it when the task needs deeper planning or analysis.” Claude Code’s says the default “suits most coding tasks.”
2. Change it in one click. Select the message box. The effort levels slide out next to the agent settings button, which shows your agent’s name. Select a level. In a narrow window, or any time, you can also open the agent settings and use the Effort column.
3. Know when it applies. The new level applies to your next message. Each conversation keeps its own effort, and new conversations start with the level you last picked for that agent.

If a request fails or you stop it, you can change the effort before retrying. A Retry with button appears next to Retry and uses your new settings.

An example

• “Change the button text on the homepage from ‘Submit’ to ‘Send’.” Low is enough. You’ll see the result in seconds and can check it yourself.
• “Add a page where customers can see their past orders.” Stay at Medium.
• “Checkout sometimes charges people twice. Find out why and fix it.” Raise it to High, so the agent checks edge cases before it answers.

Common pitfalls

• Leaving it at the top level all the time. Every message takes longer and uses more of your plan, often for little gain. See Make your agent usage last.
• Raising effort to fix a vague request. More thinking won’t guess what you meant. Make the request clearer first. See Describe what you want.
• Confusing effort with speed. Fast, in the agent settings’ Speed column, is a separate option for Claude Code and Codex. It makes supported models respond faster, and Enjoy notes that it uses paid credits or costs more.
• Comparing levels across agents. Claude Code’s documentation notes the scale “is calibrated per model,” so the same name doesn’t mean the same amount of thinking everywhere.