Which Claude plan do you need for Claude Code?
Answer four questions to see how much of Pro, Max 5x and Max 20x your usage would fill.
Max 5x fits how you work.
$100 a month. Your usage would fill about 52% of it, with room for busy days.
Pro
$20 a month
You would hit limits
2.6x the plan's limit
Max 5xBest fit
$100 a month
Room to spare
Fills 52% of the plan
Max 20x
$200 a month
Room to spare
Fills 13% of the plan
At API rates, this usage would cost about $325 a month. Max 5x gives you the same for $100.
What does “at API rates” mean?
You can also use Claude without a plan and pay for each token through Anthropic's API. Those prices are the API rates. We use them to put a dollar value on your usage. They are also what you pay for usage credits once you go past your plan's limits.
The models, in plain words
Claude Code can run on different Claude models. Bigger models handle harder work but cost more per token, so the same work uses up your plan faster. You can switch any time by typing /model in Claude Code.
Fable 5.1
The hardest and longest jobs: open-ended problems, root-cause hunts, architecture decisions
Speed: Slower
Price per million tokens it writes: $50
Uses up a plan: 2.5x as fast as Opus
Opus 5.5Default
Complex coding and long agent sessions. A strong choice for most work
Speed: Moderate
Price per million tokens it writes: $20
Uses up a plan: The baseline
Sonnet 5.5
Everyday coding when you want fast answers
Speed: Fast
Price per million tokens it writes: $10
Uses up a plan: Half as fast as Opus
Haiku 5.5
Simple, quick tasks. Rarely the main model in Claude Code
Speed: Fastest
Price per million tokens it writes: $0.50
Uses up a plan: A small fraction of Opus
What is a token? Claude reads and writes text in small pieces called tokens. Everything counts: your messages, the code Claude Code reads, and its replies. One million tokens is about 555,000 words.
How fast a model uses up a plan is our estimate, based on each model's price per token. Anthropic does not publish how each model counts against plan limits. Model overview, Claude Code model settings
Plan limits, in plain words
Plans limit how much you can use Claude in a set period of time. That is what the bars above measure.
- The 5-hour limit
- Each plan gives you a set amount of use inside a rolling five-hour window. When you reach it, you wait until the window moves on. Max 5x gives five times Pro's amount, Max 20x twenty times.
- The weekly limit
- Paid plans also have a weekly cap, which resets at a fixed time set for your account. Max plans have a separate weekly limit for Fable.
- One pool for everything
- Claude chat, Claude Code and the editor extensions all draw from the same limits. A long chat session counts too.
- When you hit a limit
- Wait for the reset, move up a plan, or turn on usage credits. Usage credits are prepaid, billed at API rates on top of your plan, and you can set a monthly spend cap.
Sources: Claude Code with Pro or Max, usage credits, plan pricing
How the estimate works
We turn your answers into what your usage would cost at API rates. Then we compare that with how much each plan covers. Anthropic publishes the ratios between plans (Max 5x gives five times Pro, Max 20x gives twenty times) but not the limits themselves, so we calibrated Pro against a real Max 20x user. Treat the result as a guide, not a guarantee.
When paying per token makes sense
For hands-on coding, a plan almost always costs less. Paying per token makes sense if you use Claude Code only a few days a month, or if it runs on its own in CI, scheduled jobs or shared tools. Those use an API key.