What Did That Prompt Cost?
See the exact dollar cost of a single prompt — and what it adds up to per day and month.
With the default inputs, $0.0225 for this prompt — Claude Opus 4.8 · $2.25 per 100 prompts · $33.75/month at 50/day. Enter your own numbers below to recompute instantly; the full step-by-step math is shown under the worked example.
Tokens served from the prompt cache (a cache hit). Most SDKs report this per request.
Results update automatically as you type.
- Fresh input tokens(1,500 @ $5.00/1M)
- $0.00750
- Output tokens(600 @ $25.00/1M)
- $0.0150
- This prompt
- $0.0225
- Per 100 prompts
- $2.25
- Per day (50 prompts)
- $1.13
- Per month
- $33.75
Ever wondered what one message to Claude, GPT or Gemini actually cost you? Paste the token counts from a single request — input, output, and (if your provider reports them) cache-read and cache-write tokens — and get the exact price of that prompt in plain dollars, then see what it becomes at 100 prompts, per day and per month. This is the money-first, per-message transparency your provider dashboard doesn't give you.
How this is calculated
Cost = (fresh input ÷ 1M × input price) + (output ÷ 1M × output price) + (cache-read ÷ 1M × cache-read price) + (cache-write ÷ 1M × cache-write price) for your chosen model. Cache-read defaults to roughly 10% of the input rate and cache-write to a premium above it; when a model publishes neither, we fall back to the base input price and say so. We then multiply by your prompts-per-day to project daily and monthly spend.
Is this a good result? What to do next
For a single chat message a few tenths of a cent is normal; long, document-heavy prompts on frontier models can run several cents each. The number that matters is the monthly projection — a cheap-looking prompt at high volume is where real money goes. If cache-read tokens dominate a big prompt, caching is already working for you.
Typical planning ranges
- Short chat message
- fractions of a cent
- Long document prompt
- a few cents on frontier models
- Cache read
- ~10% of input price
Ranges are typical planning figures to sanity-check your result, not authoritative benchmarks. Your numbers will vary with use case, volume, and vendor.
How to improve this number
- Shorten the prompt — fresh input tokens are billed at full price.
- Move stable context into a cached prefix so it bills as cache reads.
- Cap output length; output is usually the priciest per-token line.
Common mistakes
- Judging a prompt by input price alone when output dominates the bill.
- Forgetting that a cheap per-prompt cost times high volume is a large monthly number.
When to use a different approach
To compare the same prompt across every model, use the prompt cost comparison calculator. To model a whole growing session, use the agent session cost calculator.
Worked example (defaults)
With the default inputs above, here is the result:
- Fresh input tokens(1,500 @ $5.00/1M)
- $0.00750
- Output tokens(600 @ $25.00/1M)
- $0.0150
- This prompt
- $0.0225
- Per 100 prompts
- $2.25
- Per day (50 prompts)
- $1.13
- Per month
- $33.75
- cost = input + output + cache-read + cache-write, each = tokens / 1M × that rate
Sources & references
Frequently asked questions
Where do I find these token numbers?+
Every major API returns a usage object per request. OpenAI reports prompt_tokens / completion_tokens (and cached_tokens under prompt_tokens_details); Anthropic reports input_tokens, output_tokens, cache_read_input_tokens and cache_creation_input_tokens. Paste those numbers here.
Why separate cache-read and cache-write?+
Cache reads (hits) are cheap — roughly 10% of the input rate. Cache writes (creating the cache) usually cost a premium over normal input (about 1.25× on Anthropic). Splitting them gives you the true cost of a cached request instead of a rough average.
Is this per-message cost realistic for agents?+
Yes — but a single prompt is only one turn. For a whole session that grows turn by turn, use the agent session & prompt-caching cost calculator, which sums every turn.