LLM pricing · Anthropic
In short: Claude Haiku 5.5 (Anthropic, model id claude-haiku-5-5) costs $0.1 per million input tokens and $0.5 per million output tokens. A typical chat message costs about $0.00035, so a user on a $9.99 plan can send about 24,261 messages a month before they cost you more than they pay.
| Input | $0.1 |
| Cached input (cache read) | $0.01 |
| Cache write (5 minutes) | $0.125 |
| Output | $0.5 |
| Prompts over 100,000 tokens | $0.5 input, $2.5 output |
One message = 1,500 input tokens (system prompt and recent history) and 400 output tokens.
| Messages / month | AI cost / month | Margin on $9.99 | |
|---|---|---|---|
| Light user | 30 | $0.0105 | $8.48 |
| Typical user | 300 | $0.105 | $8.39 |
| Heavy user | 3,000 | $1.05 | $7.44 |
With prompt caching (two thirds of each prompt read from the cache), a message costs $0.00026, 26% less.
List prices, standard tier, prompts under the long-context threshold. Averages hide your heaviest users: FiscaLLM shows this per real user.
Averages hide the users who cost you most. FiscaLLM reads token counts from each Anthropic response, prices them with the rates above (including cache reads and long prompts), and puts each user's AI cost next to what they paid through RevenueCat, Stripe or a CSV. Free and open source, with alerts and optional per-user caps.
meter.trackAnthropic(res, { userId, feature: "chat" });See the demo dashboard or read the Claude cost per user guide.
$0.1 per million input tokens, $0.01 per million cached input tokens, and $0.5 per million output tokens (standard tier, list price).
About $0.00035 for a 1,500-token prompt and a 400-token reply, or $0.00026 when two thirds of the prompt is read from the cache.
About 24,261 a month before their AI cost uses up the $8.49 you keep after a 15% store fee. Past that, the user loses you money.
List prices from Anthropic's pricing page, standard tier; batch, priority and regional pricing differ. Check the provider for the latest. How these prices are audited.