Claude Code models and effort: /model, /effort, opusplan and fast mode
Claude Code runs on a Claude model you choose with /model, or with claude --model as you start. Sonnet suits most daily coding, Opus harder reasoning, Haiku quick simple jobs, and Fable the largest and longest tasks. /effort sets how much the model thinks, from low to max, and /fast makes Opus answer faster at a higher price.
Control
Take this page into your AI
Open in ChatGPT (new tab)Open in Claude (new tab)Claude Code is the program; the model is the Claude model it sends your conversation to. You can change the model at any time without losing the conversation, and you can tell the model how hard to think. Both choices change speed, cost and how well hard problems get solved.
Which model for which job
You choose a model by its family name, an alias that always points to the latest version Claude Code recommends:
| Alias | Use it for |
|---|---|
sonnet | Most day-to-day coding |
opus | Hard reasoning: tricky bugs, design decisions, large refactors |
haiku | Quick, simple jobs where speed matters most |
fable | The largest, longest tasks, such as an investigation you would normally split into pieces |
opusplan | Opus while planning in plan mode, Sonnet while making the changes |
default | Clears your choice and goes back to your account's default |
To pin one exact version instead of the latest, use its full id, such as claude-opus-5-5. What each model costs per token is on the price list.
Fable is the most capable, and depending on your plan it may be paid from usage credits rather than your plan's included usage; the /model picker says so on its row. It does best when you describe the result you want and let it work out the steps.
Switching
/model # open the picker
/model opus # switch at once
/model also saves your choice as the default for new sessions. To switch for this session only, open the picker and press s on the row instead of Enter. To start on a model without saving anything, use the flag:
claude --model opus
Each terminal can run its own model this way. /status shows which model you are on.
Switching mid-conversation has a cost: each model keeps its own prompt cache, so the first message after a switch reads the whole conversation again without the cache, which costs more. Switch at the start of a task rather than back and forth.
Effort: how hard the model thinks
Effort sets how much reasoning the model does before it answers. Lower is faster and uses fewer tokens; higher works through harder problems.
| Level | When to use it |
|---|---|
low | Quick back and forth where you check each result: a rename, a first sketch |
medium | Clear, well-scoped work such as building a feature. The default on Opus 5.5, Sonnet 5.5 and Haiku 5.5 |
high | Work where edge cases matter, such as fixing a bug in old code. The default on most other models |
xhigh | Deeper reasoning, more tokens |
max | Hard problems Claude works through without you. Can overthink; it applies to the current session only |
Set it with /effort high, or claude --effort high at the start. Like /model, the slider saves your choice with Enter, per model, and s applies it to this session only.
For one hard question without changing the setting, put ultrathink anywhere in the message. Phrases such as "think hard" have no special effect; they are read as ordinary words.
Long sessions: the 1 million token window
The current Opus, Sonnet, Haiku and Fable models run with a 1 million token Context WindowContext WindowEverything a model can see at once while it writes an answer, counted in tokens. When the window is full, something has to be left out. in Claude Code, enough for a long session in a large codebase. Claude Code compacts the conversation automatically shortly before it fills. To compact earlier, set the size with /autocompact, such as /autocompact 400k: smaller contexts are cheaper per message and keep Claude more focused.
Fast mode
/fast switches Opus to a configuration that answers up to 2.5 times faster, for the same quality, at a higher price per token. A ↯ next to the prompt shows it is on. If you are on another model, turning it on switches you to Opus.
It suits live debugging and rapid back and forth, when waiting costs you more than tokens. On Pro, Max, Team and Enterprise plans it is paid from usage credits, not your included usage, and Team and Enterprise need an owner to allow it. Turn it on at the start of a session rather than halfway through, for the same cache reason as switching models.
Seeing what it costs
/usage shows what the session has used and how much of your plan is left. Costs covers how to spend less, and the agent cost calculator estimates what a run costs on each model.
Questions
- Which model should I use in Claude Code?
- Start with the default for your account. Use sonnet for most day-to-day coding, opus for hard debugging, design decisions and large changes, haiku for quick simple jobs, and fable for long, open-ended tasks you would normally split up. Raise the effort level before you change model if answers feel shallow.
- How do I change the model in Claude Code?
- Type /model and pick one, or /model opus to switch at once. That also becomes your default for new sessions; press s in the picker instead to switch for this session only. To start on a model, run claude --model sonnet.
- What is opusplan?
- A model setting that uses Opus while you are in plan mode and switches to Sonnet when Claude starts making the changes, so the thinking-heavy part gets the stronger model and the typing-heavy part the cheaper one. Choose it with /model opusplan.
- What does ultrathink do?
- Putting the word ultrathink anywhere in a message asks for deeper reasoning on that one turn, without changing your effort setting. Phrases such as "think hard" are passed to the model as ordinary words.
- Is fast mode a different model?
- No. It is Opus with a setting that answers up to 2.5 times faster at a higher price per token. On a Pro, Max, Team or Enterprise plan it is paid from usage credits, not your plan's included usage. Turn it on and off with /fast.
Sources