Claude

Editorial assessment

Anthropic sizes Max as a multiple of Pro’s per-session allowance, not as a message count, so the useful threshold is how often you run out. The plan table also locks a few tools behind Pro, which settles the choice for some users regardless of volume.

• FreeFree

  is enough if you chat, search the web and make the odd file or artifact, and five projects are enough for your ongoing work.

• ProUSD 17 per month

  is the step once you need Claude Code, Research, Opus models or more projects, and you reach the Free limit only now and then.

• Max 5xUSD 100 per month

  fits if you hit Pro’s session limit several times a week; Anthropic itself suggests it for Pro users who keep hitting limits on larger code repositories.

• Max 20xUSD 200 per month

  fits if you already run out on Max 5x, or Claude is part of most of your working day.

Choose annual Pro based on the upfront commitment and expected use, not on an assumption that the unused payment disappears when upgrading. Anthropic documents credit for unused annual Pro value when moving to Max with the same billing address. Review the upgrade route and app-store terms below before committing.

On Pro and Max, the session allowance resets every five hours, and a separate weekly limit applies across all models. The weekly reset falls on a fixed day and time assigned to your account, no matter when you subscribed, and Settings > Usage shows the next one.

How many messages fit into a session depends on message length, attached files, the length of the conversation, the model and the feature you use. Chat on claude.ai, Claude Code and Claude Desktop all draw on one shared allowance, and Research sessions drain it faster because they pull in many sources.

Anthropic may also apply weekly or monthly caps, or limits on specific models and features, to manage capacity. When you run out, you can wait for the reset, upgrade, or switch on usage credits on a paid plan to keep working.

The context window sets how much a single chat can hold, and it depends on the model. On paid plans the newest models accept up to 1M tokens, others 500K or 200K, and part of the window is always kept free for Claude’s reply.

With code execution switched on, Claude summarises earlier messages when a chat nears that ceiling and keeps the full history available for reference. Those long chats use more of your allowance, so a fresh conversation is the cheaper option when you are close to a limit. For large reference sets, projects on paid plans load only the relevant parts through retrieval.

• Annual Pro is one payment made up front, not a monthly charge of the displayed monthly equivalent.
• Moving up between Max tiers is charged pro rata for the rest of the billing cycle.
• Switching from annual Pro to Max credits any unused Pro balance above the Max price to future bills, as long as the billing address stays the same.
• Subscriptions bought through Google Play are charged the full new price on upgrade, and the unused value becomes extra days on the new plan.

Prices on the pricing page exclude applicable tax, and Anthropic offers no standing discounts, only occasional limited-time promotions.

Claude’s Research runs several linked searches and returns answers with citations, and it can also draw on connected Gmail, Google Calendar and Google Docs. It is a paid-plan feature, while web search on its own is available on Free.

If most of your work is looking things up with sources, compare it with Perplexity; if you want one general assistant for many kinds of work, compare it with ChatGPT. If you would rather reach models from several companies through one account, read the Merlin AI profile and its usage terms first. Our editorial policy explains how profiles like this one are sourced and corrected.

Input: Invented policy: returns within 30 days; custom engraving excluded; damaged arrivals reported within 48 hours; contact by email. Ask for a 120-word customer explanation.

1. Request a bullet list of the rules before asking for prose.
2. Ask for a warmer rewrite that preserves every exception.
3. Compare both versions with the four original clauses and highlight any missing condition.

Accept: The 30-day rule, engraving exception and 48-hour damage window remain distinct. Stop or revise: Do not accept a rewrite that promises all items can be returned; restore the missing exception and repeat the comparison.