As listed on Claude's pricing page on 18 October 2026, Max starts at $100 a month and lets you choose 5x or 20x the usage of Pro per five-hour session. You pay for a larger usage allowance, higher output limits and priority at busy times. API use is priced separately, per million tokens.
This post uses only what the provider publishes. It makes no claims about the provider's own costs. Prices change, so confirm on the page before you budget.
What are the published plan prices?
From the pricing page, fetched 18 October 2026:
- Free: $0.
- Pro: $17 a month on annual billing ($200 up front), or $20 billed monthly. It adds Claude Code, Projects and more models, and the page says it offers at least 5x more usage per five-hour session than Free.
- Max: from $100 a month, with a choice of 5x or 20x Pro's usage, higher output limits and priority access at high traffic times.
- Team: a standard seat at $20 per seat per month billed annually, or $25 billed monthly. A premium seat is $100 annual or $125 monthly, with 5x the usage of a standard seat.
How do limits work?
The page says every plan has usage limits that reset on a rolling five-hour session window, and paid plans add weekly limits on top. It also says Anthropic may apply weekly and monthly caps or limits on model and feature usage at its discretion.
So a Max subscription is not unlimited. It is a bigger bucket, refilled on a schedule.
What happens when you run out?
The page gives three paths: wait for the reset, move to a higher plan, or turn on usage credits to keep working at standard API rates. That third option is the bridge between a subscription and metered use.
How does the API price compare?
The same page lists API prices in US dollars per million tokens. For the Sonnet 5.5 model, input is $2 and output is $10. For Opus 5.5 it is $4 input and $20 output. For Haiku 5.5, up to 100K tokens of context, it is $0.10 input and $0.50 output. The page also lists cache write and cache read prices, based on a five-minute cache lifetime.
To get a rough feel, a long coding session that sends and receives many millions of tokens will cost real money at API rates. Whether a flat plan or metered use is cheaper depends on your own volume. Count your tokens for a week before you decide.
What does this mean for security teams?
Security work comes in bursts. A quiet month followed by a week of incident response can exhaust a session limit at the worst time. Plan for that with a written fallback: who turns on usage credits, and with what cap.
You can also shrink the model's workload. Safeguard's MCP server at mcp.safeguard.sh connects to Claude, and Safeguard tool calls are unlimited on every Safeguard plan. Inventory questions, package lookups and finding retrieval run as tool calls, and the model spends its allowance on reading and reasoning.
What to do
- Record your usage for one typical week before choosing a tier.
- Pick the lowest tier that covers a normal week, not a peak week.
- Write an incident rule for turning on usage credits, with a cap.
- Move repeatable lookups to tools so they do not use model allowance.
- Compare a month of tokens against the API table to see whether metered use fits.
- Re-read the pricing page every quarter.
Quick check
Is Max unlimited?
No. The page says usage limits apply, with a rolling five-hour session window and weekly limits on paid plans.
Can I keep working after I hit the limit?
Yes, by waiting for the reset, moving up a plan, or turning on usage credits at standard API rates, according to the pricing page.
Sources
- Claude pricing (fetched 18 October 2026)
Last reviewed: 18 October 2026.
