Free calculator · runs in your browser
What will your AI app cost per month?
Not just tokens: the model API, the server, the database and email in one itemised bill. Change any input and the total updates. Share the link to keep your scenario.
Monthly bill
$26.38
$0.132 per active user
| AI modelClaude Haiku 5.5: 12,160 requests | $3.44 |
| AI modelClaude Sonnet 5.5: 3,040 requests | $16.95 |
| HostingVultr Regular Cloud Compute 1 vCPU / 1 GB | $5.00 |
| DatabaseSQLite / PocketBase / Postgres on the same server | $0.00 |
| EmailBrevo Free (300/day) | $0.00 |
| OtherDomain, monitoring, other | $1.00 |
| Total per month | $26.38 |
Sources
- platform.claude.com/docs/en/about-claude/pricing
- developers.openai.com/api/docs/pricing
- ai.google.dev/gemini-api/docs/pricing
- claude.com/pricing
- one.google.com/about/plans
- api.vultr.com/v2/plans
- hetzner.com/cloud/regular-performance/
- digitalocean.com/pricing/droplets
- digitalocean.com/pricing/app-platform
- digitalocean.com/pricing/managed-databases
- railway.com/pricing
- brevo.com/pricing/
How the estimate works
- Model: requests per day × 30.4 days, split between your two models. Each request costs (fresh input × input price) + (cached input × cache-read price) + (output × output price), using prices per million tokens.
- Hosting: the plan's monthly list price. For Railway, the Hobby fee includes $5 of usage; CPU and RAM above that are added at Railway's published rates.
- Database and email: the plan price. Running SQLite, PocketBase or Postgres on the same server costs nothing extra, which is how most of our own apps work.
Not included: tax, free credits, batch discounts, long-context surcharges (some models charge more above 100k to 272k input tokens), backups and your time. For your time, see VPS vs PaaS.
Questions
Why is the model bill often smaller than the server bill?
For a small app on a cheap model, a few hundred requests a day costs cents to a few dollars a month. A $5 server is often the biggest line. That flips quickly with a flagship model, long prompts or heavy agent use, so try your own numbers.
What is a prompt cache hit?
Most providers let you cache the fixed start of a prompt (instructions, examples, documents). Later requests that start the same way read it from cache at a fraction of the input price: about a tenth or less on the current price lists. Output tokens are never discounted by caching.
Does this include the cost of writing to the cache?
No. Writing to the cache costs more than normal input on some providers, but you pay it once per cache lifetime, not per request. For an app with steady traffic it is a small share of the bill; the calculator ignores it.
How do I find my token counts?
Most API responses report input and output tokens per request, and each provider's console shows totals per day. As a rough guide, 1,000 tokens is about 750 English words.