Using a Pricey AI Model for Easy Jobs? /model and /effort

If you're using the smartest, priciest AI model to fix a single typo, it's like taking a supercar for a trip you could make in a compact car. It's not even fast, and just costs more. Claude Code lets you switch models (/model) and even adjust thinking depth (/effort) to match task difficulty. Use just these two well and you secure cost and quality at once.

Definition

The two commands are a pair but have different roles.

• /model: Switches the AI model itself and saves it as the default. Each model differs in smartness, speed, and cost. (e.g., a fast, cheap model vs. a smart but slow model)
• /effort: Adjusts how deeply the model thinks (thinking depth). There are levels from low to max. Even with the same model, raising effort makes it more careful, lowering it makes it answer faster.

Simply put, /model sets "who works," and /effort sets "how much effort to put in."

How to use it (by difficulty)

Basics — pick a model from the list

Type into the chat box.

/model

This shows the list of usable models on screen. Pick the one you want with the arrow keys and press Enter, and it switches to that model and saves it as the default. Subsequent conversation proceeds with that model.

Applied — specify a model directly

/model opus

Append a model name directly instead of going through the list and it changes to that model immediately. (opus is the name of a model for smart tasks.) Handy for quickly stepping up right before entering a hard task.

Applied — adjust thinking depth

/effort high

Append one of low / medium / high / max after /effort. high or max ponders more deeply to raise accuracy, low answers fast. Set it high for complex design, low for simple repetition.

Advanced — change temporarily for this session only

/model

With the list shown after typing /model, pressing the s key applies only to this session and keeps the default as is. Useful when you want to "use a smart model just for this task, then back to normal next time."

Common pitfall — don't confuse model and effort

Picking a smart model with /model doesn't automatically make it think deeply. Model (who works) and effort (how much effort) are adjusted separately. For a hard problem, step up the model with /model opus and also raise the thinking depth with /effort high for it to show its true skill. Conversely, for simple tasks, lower both for speed and cheapness.

A real case

When polishing a blog post, I quickly handled simple typo and sentence fixes with a fast model at low effort. Then only at the step requiring fresh design of an SEO-conscious article structure did I step up the model with /model opus and raise the depth with /effort high. Easy stuff fast, hard stuff heavy — splitting it this way kept the result quality the same while cost dropped noticeably.

Use it like this too

• Change model directly: Specify immediately by appending a name like /model opus.
• Session only, temporarily: Press the s key after /model to apply only to this session.
• Thinking depth separately: Adjust carefulness with /effort low~max.
• Combos by difficulty: Simple tasks = fast model + low effort, hard tasks = smart model + high effort.

Tip: A pricey model isn't always the answer. 80% of tasks are fine with a fast model, and stepping up only for the truly hard 20% is the most economical.

Wrap-up

The key is one sentence. Pick model and effort to match task difficulty. Set "who works" with /model and "how much effort" with /effort. Fast for easy jobs, smart and deep for hard ones. The pricey option isn't always the right answer — turning the dial to fit the situation is the expert's cost management.

Based on: Claude Code v2.1.154 (2026.05)