Effort: How Hard It Should Think

In short

Effort is the second dial: not how much it knows, but how much it's trying right now.
It's separate from the model: you combine them however you like.
High: it reasons for a long time, takes longer, costs more. Low: it answers fast and sometimes cuts corners.
It should be chosen per job, not set once and forgotten.

Two equally good colleagues

Picture two people with the exact same background. To the first you say: «give me a quick take». To the second: «take whatever time you need and really look at it».

You'll get two different answers. Not because one of them is better — it's the same head — but because one stopped to think and the other didn't.

That's exactly the difference between model and effort. The model is how much it knows. The effort is how much effort it puts into this answer.

What actually changes

Raising the effort makes the model reason longer before it moves: it considers more paths, notices more things, makes fewer mistakes on tangled problems. Lowering it, it gets straight to the point: it answers fast and costs less.

And here's the part that surprises people: reasoning isn't free. Thinking for a long time is work, and you pay for that work twice — in waiting and in money. Keeping high effort switched on for everything is one of the quietest ways to burn through your budget.

Looping animation: the cursor clicks the «High» button in the bottom right, next to the model name, and the effort control appears: the label «High effort» above a slider that runs from «Faster» to «Smarter». — It sits next to the model, in the bottom right: first the engine's name, right after it how much effort it should put in. And it's a slider, not a switch — effort is a quantity you dial in, and the two ends are named after what you get from each.

When to raise it and when to leave it alone

Raise it when the job needs judgment: a decision you'll be living with, a problem that doesn't add up, something you need to set up right the first time.
Leave it low when the path is already clear and all that's left is to walk it: repetition, conversions, edits you've already described in full.

The rule is the same as for the model, applied to a different dial: spend the effort where there's something to figure out.

What's next

Now you have two dials, and the doubt that trips up practically everyone comes up: when Claude Code misbehaves, did I get the model wrong, or the effort? In the next lesson, the two symptoms that tell you which one to turn.

Did you get it?

Two minutes, no grade. Trying to answer — even getting it wrong — cements the concepts far more than rereading. Not in the mood? Skip it: the lesson still counts as complete.

What's the relationship between the model and effort?

What happens when you raise the effort?

1
1. 1
2. 2
3. 3
2
1. 4
2. 5
3. 6
4. 7
3
1. 8
2. 9
3. 10
4. 11
5. 12
4
1. 13
2. 14
3. 15
4. 16
5. 17
6. 18
5
1. 19
2. 20
3. 21
4. 22
5. 23
6. 24
6
1. 25
2. 26
3. 27
4. 28
5. 29
6. 30
7
1. 31
2. 32
3. 33
4. 34
8
1. 35
2. 36
3. 37
4. 38
9
1. 39
2. 40
3. 41
10
1. 42
2. 43
3. 44
11
1. 45
2. 46