Cost calculator
Estimate your monthly Claude Code spend on Opus, Sonnet high or effortless, and what AutoMode would cost, from per-step costs measured on real sessions.
At 50 questions and 20 coding tasks a week, Claude Code on Opus costs about $42.35 a month at list price, and AutoMode about $18.45, 56% less. The estimate multiplies your weekly mix by the cost of each step in our benchmark. Change the numbers to match how you work.
List price. A month here is 4.33 weeks.
On a Pro or Max plan you pay a flat price, so read the difference as the share of your usage limits you would save.
What does a switch cost inside one session?
A cheaper model does not save money from its first request. It has to write the whole chat into its own prompt cache, while the model you started on would only have read its warm cache. The saving dips first, and every cheaper request after it pays some back. AutoMode's bar shows the same thing live: ↺ $0.07 to pay back, then saved ~$0.38.
Watch a switch pay back
The same code as the bar's live saving, on an example chat. Real prices and cache rules; example token counts.
Example chat: 1,500 new tokens and 800 output tokens per request, 1-hour cache, one request a minute. AutoMode's judges are left out here; the bar counts them.
The chart runs the same code as the bar, hooks/lib/savings.ts. We checked that code against the benchmark by replaying its five mixed sessions: it estimated a 56% saving against Opus, and the measured Opus runs showed 52%. Read the calibration.
Where do the numbers come from?
Each arm ran the same five mixed sessions on real repositories: a question, a task, another question, then a harder task, in one conversation. Hidden tests graded every task, and every arm passed all 20 steps. Costs are list-price equivalents of every paid request, judges and subagents included.
| Cost per step | Question | Coding task |
|---|---|---|
| Claude Code on Opus (the default) | $0.050 | $0.364 |
| Sonnet high | $0.024 | $0.190 |
| effortless, on Opus | $0.043 | $0.344 |
| AutoMode | $0.008 | $0.193 |
When is the estimate wrong?
- Your tasks are longer or shorter than the benchmark's. Cost grows with context, so long sessions cost more on every arm.
- You only run coding tasks on Sonnet high. Then AutoMode costs about the same, a little more for its judges and checks.
- One repetition per arm. The benchmark is a release signal, not a law. Read its method and limits.
Install AutoMode
Paste these into Claude Code, one at a time, then run /reload-plugins. It is free, needs no key and no server, and starts on its own.
/plugin marketplace add mralexivy/AutoMode
/plugin install automode@automodeLast updated on
Delegated reads
AutoMode's optional read gate: a full read of a large file stops once with a free outline and a cheap Haiku extract tool. How it works and stays safe.
How to reduce Claude Code costs
Practical ways to cut Claude Code token use and spend: match the model to the job, lower effort, clear context, keep the cache warm, and route each prompt.