Skip to content
Book a callSkip the hiring grind

Day 27 of 52 · about 15 min · Basics

Choose a model and effort, and watch your usage

Running every task on the biggest model at the highest effort feels safe, but it is slower and burns through your plan or budget. Matching both to the task, and glancing at /usage, keeps you quick and under your limits.

You’ll finish with

  • A default model and effort, set with /model and /effort
  • A read of /usage: where your usage went
  • Three habits that spend fewer tokens

You’ll need

  • Claude Code installed and logged in
  • Optional: fast mode (/fast) is a research preview and, on subscription plans, bills to usage credits only.
Start from the practice project as of Day 23 (.zip, 30 KB)

Before you start

Tell us about your setup and the steps change to fit it. Your answers stay in this browser, and you can change them at any time.

Which system are you on?

Commands differ between systems. Pick the one you’ll run Claude Code on.

The steps

One step at a time. When a step’s check matches, tick it and the next one opens.

0 of 5 steps done0%

  1. Step 1 of 5: See what you run on now

    Open the model picker and decide how long a choice should last.
    In Claude Code

    You type: /model

    What happens: Enter switches and saves it as your default; s switches for this session only.

    With no argument, /model opens the picker.

    Check it worked

    The session header shows your model and its effort, for example "with low effort".

  2. Step 2 of 5: Match the model to the task

    Use the alias that fits the job. Aliases update over time.

    sonnet

    • Most coding tasks
    • Costs less than Opus

    opus

    • Architecture decisions
    • Multi-step reasoning

    haiku

    • Simple tasks
    • Simple subagents, with model: haiku
    opusplan plans with Opus, then builds with Sonnet.
    In Claude Code
    /model sonnet

    Each model has its own cache, so switch at the start of a task, not mid-conversation.

    Check it worked

    The header names the model you picked.

  3. Step 3 of 5: Set the effort level

    Match the effort level to the task.

    low

    • Quick changes you review, like a rename

    high

    • Bug fixes where edge cases matter

    max

    • Hard problems, for this session only
    Levels depend on the model. Organizations can cap them.
    In Claude Code
    /effort high

    Note: For one hard question, add ultrathink to the prompt without changing your setting.

    Check it worked

    /effort status shows the level in effect.

    If it didn’t work

    The level in effect isn’t the one I set

    The model may not support it, so Claude Code uses the highest level at or below it. An organization cap also wins.

  4. Step 4 of 5: Read /usage

    See what you used, and what used it.
    In Claude Code

    You type: /usage

    What happens: Session block: tokens and an estimated cost, mainly for API keys.

    What happens: Subscriptions: plan bars and each skill, subagent, plugin and MCP server’s share. d or w picks 24 hours or 7 days.

    The breakdown counts this machine only.

    Check it worked

    You can name the largest share in the breakdown, or the session’s token total.

  5. Step 5 of 5: Spend fewer tokens

    Keep context small with three habits from the docs.

    Uses more

    • One long session for unrelated tasks
    • Workflow steps kept in CLAUDE.md
    • MCP servers you never use

    Uses less

    • /rename, then /clear between tasks
    • The same steps in a skill, loaded when needed
    • Those servers turned off in /mcp
    CLAUDE.md loads every session. Skills load only when used.

    Check it worked

    After /clear, the Session block in /usage starts again from zero.

Finish · 5 of 5 steps to go

You choose model and effort on purpose

Check /usage weekly, and change your defaults when the work changes.

What you’ll have checked

  • The header shows your chosen model and effort
  • /usage shows where your usage went

Lesson updates by email

Get lesson updates by email

Leave your email and we’ll tell you when we add or update a lesson.

We’ll email you when we add or update a lesson. We use your email only for this. Privacy policy

Claude Code for your team

Want Claude Code set up across your team?

HorizonLux sets up and repairs Claude Code for teams: CLAUDE.md rules, hooks that enforce them and reviewed MCP servers. Book a free call to talk it through.

Book a free call Claude Code agency

Related

Free tool

LLM Pricing Comparison

Free tool

Prompt Caching Calculator

How we build

AI software factory

Questions

Questions about Day 27

Should I turn on fast mode?

Only when speed matters more than cost. It runs Opus faster, at a higher price per token.

Sources

Checked against the official Claude Code docs on .

Independent resource, not affiliated with Anthropic. HorizonLux wrote this lesson; Claude Code is made by Anthropic.