Skip to main content

FREE · PRIVATE · BUILT FOR REAL WORKLOADS

Every token has a price. Every task has a real cost.

The LLM cost calculator for decisions beyond a price table. Compare monthly API spend and cost per accepted task with your own workload, cache usage, and quality requirements.

Calculate my costs Compare Haiku & Luna

No account. No API key. Your calculations stay in your browser.

Your workload. Two models. One clear comparison.

Start with the LLM cost calculator below. Change the demo numbers to match your API usage; results update as you type.

USD · Standard text APIPrices checked: 2026-10-12 ↗
Try a demo

Illustrative inputs, not industry averages. A preset replaces request rows only.

Request class 01

Claude Haiku 5.5Anthropic

Short-input tier · 2,000 tokens

GPT-6 LunaOpenAI

Short-input tier · 2,000 tokens

No caching included in this estimate.

Custom rates

Your estimate

Official Standard prices

Claude Haiku 5.5

Estimated monthly cost

$13.50

30,000 requests / month

Per request
$0.000450
Per day
$0.45
Per 1,000 requests
$0.45
Monthly cost breakdown
Uncached input tokens
$6.00
Cache-read tokens
$0.00
Cache-write tokens
$0.00
Billed output tokens
$7.50
Extra cost / request (USD)
$0.00

GPT-6 Luna

Estimated monthly cost

$13.50

30,000 requests / month

Per request
$0.000450
Per day
$0.45
Per 1,000 requests
$0.45
Monthly cost breakdown
Uncached input tokens
$6.00
Cache-read tokens
$0.00
Cache-write tokens
$0.00
Billed output tokens
$7.50
Extra cost / request (USD)
$0.00
Equal estimated monthly cost for this workload.

The difference comes from the entered usage, cache settings, extra charges, and applicable tiers. It does not establish which model performs better.

Before taxes and credits. Batch, tools, regional fees, and other unentered charges are not included. Pricing sources & methodology ↗

A clearer picture of your AI budget

The smallest token price is only the start of a useful comparison.

See where your monthly spend actually goes

Use the LLM cost calculator to separate input, output, cache reads, and cache writes. Add a row for each request length so a few long documents do not disappear inside an average. Every model keeps its own token counts.

Compare API pricing See the calculation method
Two streams of token tiles flowing into separate cost receipts

Pay attention to the tasks you can actually use

An inexpensive response can still need a retry or human review. Add observed task counts and the full cost of your sample to compare accepted outcomes. No invented quality scores: the LLM cost calculator works with your evidence.

Calculate accepted task cost Read the practical guide
Task cards pass through a quality gate, with a loop for retries

From a guess to a useful estimate

Three steps to make this LLM cost calculator reflect your application.

  1. 01

    Start with a representative workload

    Choose a demo for classification, summaries, support, or extraction. Replace its token counts and daily volume with your usage. Demo presets are examples, never industry averages.

    • Use billed output, including reasoning
    • Keep long requests in a separate row
  2. 02

    Make the billing assumptions explicit

    Open cache and extra costs when your workflow uses them. Review the official price sources, or enter a clearly marked custom rate for a contract or another provider.

    • USD, Standard text pricing
    • Tax, tools, and regional fees need review
  3. 03

    Check the cost of an accepted outcome

    Run the same task sample on both models outside this website. Enter total spend and final accepted tasks once. The LLM cost calculator will show the tradeoff without applying retries twice.

    Read the methodology

Built around the work you ship

One LLM cost calculator, four practical starting points. Each preset is fully editable.

Product classification

Estimate large labeling jobs. Accept an answer only when its category matches your labeled evaluation set.

Try classification

Document summaries

Account for long input and billed reasoning. Check that every required fact survives the summary.

Try summarization

Customer support

Budget response drafts, including repeated context. Define correctness against your support policies.

Try support replies

Structured extraction

Estimate document-to-JSON workflows. Test both schema validity and whether extracted values are correct.

Try extraction

LLM cost calculator questions

Clear assumptions make a useful estimate.

What does this LLM cost calculator include?

Standard text input and billed output, separate cache reads and writes, and extra per-request costs you explicitly add. Monthly mode estimates API calls. Accepted-task mode uses your observed total cost and final accepted tasks.

Why can the same token prices lead to different bills?

Models can use different input and output token counts, and requests can cross different long-input thresholds. Cache writes and retries also change the total. Compare your real workflow, rather than assuming identical usage.

Does the LLM cost calculator measure model quality?

No. You provide the evaluation sample and acceptance standard. The tool checks your observed acceptance rate and optional latency against your requirements; it does not run models or create benchmark scores.

Can I save or share an estimate?

Save a plan in this browser, load it later, or copy a link containing its numeric assumptions. Shared parameters live in the URL fragment. Anyone with the link can read those numbers, so only share non-sensitive plans.

Why might my invoice be different?

Your invoice may include tools, regional premiums, tax, credits, other processing modes, or a different mix of requests. Check the linked methodology and current provider prices before treating the estimate as a budget.

Make your next model decision with numbers.

Bring your workload to the LLM cost calculator.

See the monthly bill, then check what a useful result really costs.

Estimate my workload Read the getting-started guide

Free to use · Independent of Anthropic and OpenAI