AI API Cost Calculator
Estimate your monthly OpenAI, Anthropic, or Google Gemini API spend. Switch between single-call, multi-step agent, and reverse-budget modes — and toggle the OpenAI Batch API discount. Free, no sign-up.
Prices sourced from OpenAI, Anthropic, and Google official pricing pages · Last updated: April 2026
Your usage
Multimodal model with vision — text, images, and audio in one API
Async jobs, typically processed within 24h. Not for realtime UX.
Total API calls across all users per day.
System prompt + user message. 1 token ≈ 0.75 words.
Output costs are usually 3–5× input.
Estimated cost
Monthly estimate
$187.50
$6.25/day · $0.0063/request
If this workload can run async (e.g. nightly summaries), OpenAI Batch API drops it to $93.75/mo — save $93.75.
Cost breakdown
Monthly requests
30,000
Total monthly tokens
30,000,000
Input price
$2.50/M tokens
Output price
$10.00/M tokens
This is an estimate based on standard pricing. Actual costs may vary with prompt caching, volume commitments, or context-window surcharges. Verify against your provider's dashboard.
This number assumes every customer looks average. They don't.
In every product we've looked at, the top 10% of customers generate 4–10× the token volume of the median. PerUnit shows you who they are — and which features and pricing tiers they sit in — so the bill above stops being a single line and becomes a decision.
Get early access to PerUnitFrequently asked questions
- How do I estimate the cost of an AI agent or multi-step workflow?
- Switch this calculator to 'Agent run' mode. A multi-step agent typically does 4–12 LLM calls per task — one for planning, several for tool calls, one for the final answer. Most teams underestimate agent cost by 5–10× because they only count the visible call, not the planning and tool-use steps in between.
- How much can OpenAI Batch API save me?
- Batch is a flat 50% off input and output rates for jobs that can finish within 24 hours — overnight summarisation, embeddings, evaluations, content scoring. Toggle the Batch API checkbox above when an OpenAI model is selected to apply the discount to the estimate. Batch isn't available on Anthropic or Google.
- Can I work backwards from a monthly budget?
- Yes — switch to 'Monthly budget' mode. Enter what you're willing to spend, your average input and output tokens per request, and the calculator returns how many requests per month and per day that budget covers on the selected model.
- How does Claude pricing compare to GPT-4o here?
- Claude Sonnet 4.6 ($3.00/M in, $15/M out) sits roughly alongside GPT-4o ($2.50/M in, $10/M out) at the flagship tier — Claude is a touch more expensive on output. Claude Haiku 4.5 ($1.00/M in, $5/M out) is the closest analogue to GPT-4o mini, but mini is still meaningfully cheaper on input. Switch the model selector above to see your numbers at each.
- How accurate is this estimate?
- It uses standard-tier pricing from OpenAI, Anthropic, and Google. Real bills run lower with prompt caching, batch processing, or negotiated volume discounts — and higher if you exceed certain context-window thresholds on Gemini. Use this as a planning number, not a substitute for your provider's billing dashboard.
Get a monthly email when AI model pricing changes
One email a month. New model launches, price cuts, deprecations across OpenAI, Anthropic, and Google. No spam, unsubscribe any time.