A worked estimate you can check line by line. Change the numbers to your own in the cost calculator.
What we assume
A system prompt of about 1,500 tokens plus earlier turns of the conversation (about 4,000 tokens on an average turn), then about 500 new tokens (the user's message and the last reply) and a 400-token answer. An active user sends about 300 messages a month.
Cost per message
| Model | No caching | With caching | Per 1,000 messages, with caching |
|---|---|---|---|
| Claude Fable 5.2 | $0.052 | $0.022 | $21.80 |
| Claude Opus 5.5 | $0.026 | $0.011 | $11.30 |
| Claude Sonnet 5.5 | $0.013 | $0.0057 | $5.65 |
| Claude Haiku 5.5 | $0.00065 | $0.00030 | $0.303 |
Cost per month
| Model | One active user (300 messages/month) | 1,000 active users |
|---|---|---|
| Fable 5.2 | $6.54 | $6,540 |
| Opus 5.5 | $3.39 | $3,390 |
| Sonnet 5.5 | $1.70 | $1,695 |
| Haiku 5.5 | $0.091 | $90.75 |
Where the money goes
- Conversation history is the cost that grows quietly: every new message re-sends the earlier turns. Caching makes those re-sent turns cheap, and trimming or compacting long conversations keeps them from growing without limit.
- Most chat traffic does not need the largest model. Many products use a smaller model by default and switch to a larger one only for hard requests.
- Charge or cap per user. A few heavy users can cost more than everyone else combined, so put a monthly dollar limit on each account.
To check what a plan price leaves you after these costs, use the AI SaaS margin calculator. To see how much caching saves on your own prompt sizes, use the prompt caching savings calculator.
Questions
How much does one message cost with the Claude API?
With prompt caching, from about $0.00030 on Claude Haiku 5.5 to about $0.022 on Claude Fable 5.2, for a message of 4,000 reused + 500 new input tokens, 400 output tokens. Without caching it is $0.00065 to $0.052.
How much does it cost per month?
One active user (300 messages/month): about $0.091 on Haiku 5.5 up to $6.54 on Fable 5.2. 1,000 active users: about $90.75 on Haiku 5.5 up to $6,540 on Fable 5.2, with caching.
How can I make it cheaper?
Conversation history is the cost that grows quietly: every new message re-sends the earlier turns. Caching makes those re-sent turns cheap, and trimming or compacting long conversations keeps them from growing without limit.