Claude's pricing confuses a lot of people. Here's the simple version, and the handful of habits that cut most bills in half.
If you use claude.ai on Pro or Max, you pay a flat monthly fee and have a usage limit that resets. If you use the API or Claude Code, you pay per token — roughly ¾ of a word — for everything you send and everything Claude writes back.
Output tokens cost roughly five times more than input tokens. So the length of Claude's answer matters far more than the length of your question. 'Be concise' is genuinely a cost lever, not a style note.
Claude has no memory between messages, so your whole conversation is re-sent on every turn. A long chat pays for all of itself again each message. Start a fresh chat per topic and you'll cut a surprising amount.
The smallest models cost a fraction of the largest and handle most everyday tasks fine. Reserve the top model for genuinely hard reasoning.
We built a free calculator that shows your actual monthly and yearly spend from your usage, plus an itemised breakdown of what you'd save. Most people are surprised.