Customer support chatbot on Claude Haiku 5.5: what it costs
Customer support chatbot: 3,000 requests a day, each with 2,780 input tokens (54% of it read from the prompt cache) and 250 output tokens. On Claude Haiku 5.5 that comes to about $24.46 a month, or $0.268 per 1,000 requests, at list prices from Anthropic.
It ties with GPT-6 Luna as the cheapest of the 12 models priced for this use case; the next, Gemini 3.5 Flash-Lite, costs $96.18 a month. Without prompt caching the same traffic would cost $36.77 a month. Output tokens make up 47% of the bill.
Prices on this page last verified on Sources
Where the money goes
| Token type | Tokens a month | Price per million | Cost a month | Share |
|---|---|---|---|---|
| Input | 116,800,000 | $0.10 | $11.68 | 48% |
| Cached input | 136,875,000 | $0.01 | $1.37 | 5.6% |
| Output | 22,812,500 | $0.50 | $11.41 | 47% |
| Total | $24.46 |
Price tier used: prompts up to 100,000 input tokens.
Token assumptions
A chat assistant on a help site that answers questions using a fixed policy prompt and the conversation so far.
| Part of the prompt | Tokens | Cached | Why |
|---|---|---|---|
| Instructions and support policies | 1,500 | Yes | Tone rules and the policies the bot must follow. Identical on every request, so it can be read from the prompt cache. |
| Conversation so far | 1,200 | No | Earlier turns are sent again with each request. This is the average over a five-turn conversation. |
| Customer message | 80 | No | A typical question of one or two sentences. |
| Total | 2,780 | 54% cached |
- Requests a day: 3,000
- About 600 conversations a day of five turns each. Each turn is one request.
- Output tokens per request: 250
- A short answer of about 180 words.
- Batch API: not used
- Replies are needed in real time, so nothing goes through a batch API.
These are stated assumptions for illustration, not measurements. Cached parts are assumed to be read from the prompt cache on every request. Token counts are the same for every model, although providers count tokens differently. Use the calculator to try your own figures.
The same use case on every model
| Model | Rank | Per 1,000 requests | Per month | vs Claude Haiku 5.5 |
|---|---|---|---|---|
| Claude Haiku 5.5 | 1st | $0.268 | $24.46 | |
| GPT-6 Luna | 1st | $0.268 | $24.46 | 1× |
| Gemini 3.5 Flash-Lite | 3rd | $1.05 | $96.18 | 3.9× |
| Mistral Large 4 | 4th | $1.53 | $139.25 | 5.7× |
| Gemini 3.8 Flash | 5th | $2.01 | $183.41 | 7.5× |
| GPT-6.1 Terra | 6th | $2.61 | $237.71 | 9.7× |
| Claude Sonnet 5.5 | 7th | $5.21 | $475.41 | 19× |
| Gemini 4 Argon | 7th | $5.21 | $475.41 | 19× |
| GPT-6.1 Sol | 7th | $5.21 | $475.41 | 19× |
| Claude Opus 5.5 | 10th | $10.42 | $950.83 | 39× |
| Claude Fable 5.2 | 11th | $20.54 | $1,874 | 77× |
| GPT-6 Astra | 12th | $26.80 | $2,446 | 100× |
Not priced for this use case
- Mistral Small 4: No verified cached input price for this model yet.
Your traffic looks different? Enter your own figures in the calculator.
Related pages
Customer support chatbot: other models
Sources
Sources: platform.claude.com (checked )
Prices are list prices as each provider publishes them, used unchanged. How the figures are calculated: methodology.