The Claude Code /model command switches which Claude model you are using and saves it as your default for new sessions. Claude comes in a few tiers -- some are slower but far more capable, others are fast and cheap -- and this one command lets you pick the right one for the job in front of you. Think of it like choosing between a freight truck and a scooter: both move things, but you match the vehicle to the load.

What the /model command does

When you run /model, Claude Code switches the active model for your work. Run it on its own, with no model name after it, and it opens a picker so you can choose from a list. Run it with a model name and it switches straight to that one. Either way, the choice sticks: it becomes your default for new sessions, so you are not re-picking every time you sit down.

/model

Run it with no name to open the picker. Click to copy.

Inside the picker you can also switch for just this one session instead of changing your default. Highlight a model row and press s, and Claude uses it for the current session only, leaving your saved default untouched. If the conversation already has output in it, the picker asks you to confirm before it swaps, because the next reply re-reads the whole conversation from the new model.

The three tiers, in plain English

You do not need to memorize version numbers. What matters is the trade-off between capability, speed, and cost. Here is how the tiers stack up.

model tiers
OpusThe most capable. Best for hard reasoning and complex, multi-step work. Most capable
SonnetThe balanced, everyday workhorse. A strong middle ground for routine work. Balanced
HaikuThe fastest and cheapest. Good for quick or simple tasks. Fastest

More capable costs more and runs slower. Match the tier to the task.

When to pick each model

The whole point of switching is that you are not stuck paying for a heavyweight on light work, or asking a lightweight to do heavy lifting. A simple way to decide: start from the difficulty of the task and move up only when you need to.

QuickHaikuFast lookups, short answers, simple edits. You want speed and low cost.
RoutineSonnetEveryday work that needs solid judgment but is not unusually hard. The default for most days.
HardOpusGnarly problems, complex reasoning, work where a wrong turn is expensive.

A real example

Say you are working through a messy piece of business logic that touches several files and has to be exactly right. That is Opus territory, so you switch up to it and let it do the heavy thinking:

/model opus

Once the hard part is done and you are back to routine edits and cleanup, there is no reason to keep paying for the heavyweight. You drop back to Sonnet for the everyday work. And later, when you just need a fast fact check or a quick lookup, you switch to Haiku to save time and cost. Same session, three different tools, each matched to what the moment actually needs.

Note: Switching for one session only is easy to miss. In the picker, press s on a model to use it for the current session without changing your saved default. Handy when you want to reach for Opus on one hard task and then go right back to your normal setup.

How /model relates to /effort

There is a companion command, /effort, that controls something different. Where /model chooses which Claude you are talking to, /effort sets how much reasoning effort that model spends, on models that support it. It runs from low up to max. So you can pick a capable model and still dial its effort down for lighter work, or turn effort up when a problem needs more careful thinking. The two work together: model is the tier, effort is how hard it leans in.