1 · Your tool
Using Gemini, Copilot or another tool? The advice still works. Most tools have a small, middle and top model, so match the tier you get here. Artificial Analysis lists current models by maker.
2 · The job
What are you asking it to do?
Anything else true? Pick one, or neither.
The three models
Match the model to the job, not the biggest one for everything
The biggest models
Why the biggest models aren't in the picker
- They cost a lot more per job. On Artificial Analysis tests (Nov 2026), they ran up to about five times the cost of the Top model for the same task.
- The gain is small for business work. They pull ahead on hard, multi-hour reasoning. For emails, numbers and plans, Top gets you nearly the same answer.
- They're slower and eat your limits. Some take nearly twice as long, and most sit on a higher plan.
- When to reach for one: Top on High effort still falls short on a hard problem, and the answer is worth the wait.
The second dial
Effort is how long it thinks
Low
Answers fast. Right for clear, simple asks.
Medium
A little thinking. Right for most everyday work.
High
Thinks it through. Slower, and uses your limits faster.
Good habits
Five habits that save time and limits
- Start on Everyday. Move down for simple pulls and up for real analysis.
- Too shallow? Turn effort up first. Move up a model only if that doesn't fix it.
- Too long? Ask for shorter. Lower effort won't shorten the answer.
- Context beats a bigger model. A clear ask with the right files does more than switching up.
- Names change, the three levels don't. If a name here isn't in your menu, pick the one in the same spot.