How is the monthly cost calculated?

For each model: monthly = (messages × input tokens ÷ 1,000,000 × input price) + (messages × output tokens ÷ 1,000,000 × output price). Input and output are priced separately because output is usually several times more expensive.

Are the prices accurate?

They reflect published list prices per million tokens and are labelled as estimates. Providers change pricing often and may offer volume discounts, so confirm the live rate in your provider dashboard before budgeting.

How do I estimate tokens if I only know word counts?

As a rough rule one English word is about 1.3 tokens, so multiply your average word count by 1.3. For exact counts use a model-specific tokenizer.

Is my data sent anywhere?

No. The entire calculation runs in your browser. Nothing you enter is uploaded, stored or logged.

LLM Pricing Calculator

Know your AI bill before you ship

Token pricing is easy to underestimate — a feature that feels cheap per request can cost thousands at scale. This calculator projects your monthly spend across ten leading models from one set of inputs, so you can pick the model that fits your budget instead of discovering the cost on your first invoice.

How it works

You provide three numbers: monthly request volume, average input tokens, and average output tokens. For each model the calculator computes input and output cost separately — providers charge different rates for the prompt you send and the completion they generate — using published list prices per million tokens:

monthly = (requests × input_tokens  / 1e6 × input_price)
        + (requests × output_tokens / 1e6 × output_price)

The results are ranked cheapest first, with the lowest-cost model highlighted, so the cost spread between a small-flagship model and a premium reasoning model is immediately visible — often a 50-100x difference for the same workload.

Tips to cut your bill

Right-size the model. Most tasks do not need a premium model; GPT-4o mini or Gemini 1.5 Flash often deliver the result at a fraction of the cost.
Trim output tokens. Output is the expensive side. Ask for concise answers and cap max_tokens.
Cache and dedupe. Prompt caching and reusing results for repeated inputs cut input cost on high-volume pipelines.
Re-run the calculator whenever your token estimates change — small per-request differences compound fast at scale.