Claude Pro vs Max for Claude Code

The three plans, side by side

These are Anthropic's published ranges. The spread inside each range is not vagueness; it's the difference between a short prompt on a small file and a long agentic session on a large repo. The same plan gives a light user four times the prompts it gives a heavy one, because every request resends your whole conversation and the limits are measured in tokens, not prompts.

Two limits apply at once: a 5-hour window that resets five hours after your first message, and a weekly cap that resets on a rolling seven-day cycle. On Max plans there is also a separate weekly Opus allowance. Hitting any of them pauses Claude Code and shows the reset time. The weekly limit guide covers how they interact.

What Pro actually buys

Pro is the plan most people should start on and many should stay on. At the published 40–80 Sonnet hours a week, it covers a normal workday of coding assistance if you're not running agents. In practice, Pro fits you if you:

• Use Claude Code for one to three focused sessions a day, an hour or two total.
• Work in one terminal, not three.
• Are fine with Sonnet. Pro does not include Opus in Claude Code (it does in chat).
• Clear context between tasks (/clear), which is the single biggest lever on how far a 5-hour window stretches.

The tell that you've outgrown it is the 5-hour message. Hitting it once a week is normal. Hitting it most days, before lunch, means your sessions are heavier than Pro is sized for, and the fix is either the habits in the token-saving guide or the next plan up.

What Max 5x buys

Max 5x is where full-time use starts. Five times Pro's allowance is roughly 140–280 Sonnet hours a week, which is more than a 40-hour week even at the low end, plus 15–35 hours of Opus. Anthropic's Claude Code plan page notes that on Max 5x, Claude Code automatically switches from Opus to Sonnet once you've used about 20% of your usage limit, so Opus is a budget within the budget rather than a default. Max 5x fits you if you:

• Have Claude Code open most of the day.
• Want Opus for planning and architecture, Sonnet for the rest.
• Run a subagent or a second session sometimes, not constantly.
• Also use Claude chat or Cowork heavily on the same account (they share the limits).

The plan calculator turns hours per day and session intensity into an estimate of where you land in each range.

What Max 20x buys

Twenty times Pro sounds like it should be the "never think about it" plan, and for one person in one terminal it is. But the people who buy it are usually not one person in one terminal: they're running parallel Claude Code sessions, agent teams, or automation that fires all day. Anthropic notes that agent teams use roughly seven times the tokens of a standard session, so a 20x plan feeding three teammates is a 20x plan being used like a 7x one. On Max 20x the automatic Opus-to-Sonnet switch happens at about 50% of usage instead of 20%, which is the practical difference for Opus-heavy users: more of your week can be Opus before you're nudged down.

Max 20x fits you if Max 5x's weekly cap is a real constraint, if you want Opus as your default model, or if you run several sessions at once. If none of those is true, the extra $100 buys headroom you won't use.

When the API beats all three

Anthropic publishes a benchmark for Claude Code on the API: about $13 per developer per active day, $150–250 per month, with 90% of users under $30 a day. That's more than Max 5x for anyone who codes daily, so the subscription is the cheaper way to buy the same tokens. The API wins in three situations: you use Claude Code a few days a month (the bill is a few dollars, not $20); you need it running unattended in CI or on a server, which the subscription isn't for; or you need no rate limits at all and are willing to pay for them. The API cost calculator does the per-token math with caching included.

A middle path exists: stay on Pro or Max and turn on usage credits. When the plan runs out, extra usage bills at standard API rates up to a monthly limit you set, and Claude Code prompts you in the terminal when you reach it. It's the cheapest way to survive a heavy week without upgrading for the whole month.

Team plans, briefly

Claude for Teams sells Standard seats at $25/month ($20 annual) with more usage than Pro, and Premium seats at $125 ($100 annual) with five times the Standard allowance and Claude Code included. For a company, a Premium seat is Max 5x with admin controls, a spend report, and per-member usage-credit limits. For an individual, Max is the same money with no admin.

How to choose in one minute

1. Count hours per weekday you have Claude Code actively working, not open.
2. Under 2 hours, one terminal, Sonnet is fine: Pro.
3. 2–6 hours, or you want Opus, or you use chat heavily on the same account: Max 5x.
4. 6+ hours, parallel sessions, agent teams, or Opus by default: Max 20x.
5. A few days a month, or unattended automation: API.

You can change plans any time, and Claude Code's /usage command shows your position in both the 5-hour and weekly windows, so the honest approach is to start one tier lower than you think and watch the bars for a week.

Run the plan calculator with my numbers →

Frequently asked questions

Sources

• Anthropic: Claude plans and pricing (Pro, Max, Team, Enterprise).
• Anthropic support: Using Claude Code with your Pro or Max plan (prompts per 5 hours, weekly hours, Opus switching).
• Claude Code docs: Manage costs effectively (API benchmarks, usage credits, /usage).
• Anthropic support: Extra usage (usage credits) for paid Claude plans.