LLM Token Cost Calculator

Frequently Asked Questions

How is LLM API cost calculated?

Cost = (prompt tokens × input price) + (completion tokens × output price), where prices are quoted per 1K or per 1M tokens. At $3 per 1M input and $15 per 1M output, 2,000 in + 500 out costs 2,000 × 3 ÷ 1e6 + 500 × 15 ÷ 1e6 = $0.0135.

Why are output tokens usually more expensive?

Generating each output token requires a full forward pass, while input tokens are processed in a single parallel prefill, so providers typically price output 2–5× higher than input. Trimming verbose responses saves the most money.

How do I estimate monthly cost at scale?

Multiply per-request cost by requests per day × 30. One million requests/month at $0.0135 each is about $13,500/month, so small per-token savings compound quickly.

Roughly how many tokens is my text?

For English, 1 token ≈ 4 characters ≈ 0.75 words, so 1,000 words is roughly 1,300 tokens. Code and non-Latin scripts tokenize less efficiently, so use the model's own tokenizer for billing-accurate counts.

Provided by AllCalculators.io
Free online calculators for everyday. No registration required.

Estimates for informational purposes only.

Important Disclaimer: Estimates for informational purposes only.

This calculator provides estimates for informational purposes only. Results are based on assumptions and may not reflect actual outcomes. Consult qualified professionals in relevant fields before making important decisions based on these results.

JavaScript is required to use the interactive calculator above. The questions and answers below remain readable without JavaScript.

How This Calculator Works

Large-language-model APIs bill by the token - a token is roughly three-quarters of an English word, or about four characters. Crucially, providers charge two different prices: a lower rate for input tokens (your prompt, system message, and any context you send) and a higher rate for output tokens (the model's generated completion). Prices are usually quoted per 1,000 or per 1,000,000 tokens. This calculator takes your average prompt and completion token counts, the per-1K or per-1M input and output prices for the model you are using, and an optional requests-per-day figure, then returns the cost of a single request and the projected daily, monthly, and yearly spend. It is the tool for budgeting a feature before you ship it, comparing two models, or working out whether a longer system prompt is worth the recurring cost across millions of calls.

Formula / Method

Per-request cost = (promptTokens ÷ divisor) × inputPrice + (completionTokens ÷ divisor) × outputPrice, where the divisor is 1,000 if prices are quoted per 1K or 1,000,000 if quoted per 1M. Splitting the two rates matters because output is typically 3–5× the input price, so a chatty completion can dominate the bill even with a long prompt. Recurring spend scales linearly: daily = perRequest × requestsPerDay, monthly = daily × 30, and yearly = daily × 365. We use a flat 30-day month for quick estimation; a precise monthly figure would use the actual day count. All token counts are the average per request - real traffic varies, so treat the projection as a planning midpoint, not a guarantee.

Worked Example

Suppose a model costs $3 per 1M input tokens and $15 per 1M output tokens, and your average request sends 1,500 prompt tokens and receives 500 completion tokens. Input cost = (1,500 ÷ 1,000,000) × $3 = $0.0045. Output cost = (500 ÷ 1,000,000) × $15 = $0.0075. Per request that is $0.0045 + $0.0075 = $0.012 - just over a cent, but notice the 500 output tokens cost more than the 1,500 input tokens because output is 5× the rate. At 1,000 requests per day that is $12.00/day, about $360/month, and roughly $4,380/year. Trim the system prompt by 500 tokens and you save (500 ÷ 1,000,000) × $3 × 1,000 × 30 = $45/month - small per call, real at scale.

Common Mistakes

Using one price for both directions - output tokens usually cost several times more than input tokens.

Confusing per-1K and per-1M pricing - a 1,000× error; always match the divisor to how the price is quoted.

Counting only the visible user message - the system prompt, prior turns, and retrieved context are all billed input tokens.

Forgetting that conversation history re-sends earlier turns each call, so input tokens grow as a chat gets longer.

Ignoring prompt caching and batch discounts - many providers offer big reductions that this raw estimate does not include.

Tips & FAQ

How many tokens is my text? Roughly words × 1.3, or characters ÷ 4 for English; code and other languages tokenize differently, so use the provider's tokenizer for precision. Why is output so expensive? Generating tokens one at a time is more compute-intensive than reading the prompt in parallel, so providers price it higher. Can I cut costs? Shorten system prompts, cap max_tokens, use a cheaper model for easy calls, and enable prompt caching for repeated context. Does streaming change the price? No - you pay for the same tokens whether streamed or returned at once. What about embeddings? Those are priced per input token only and are far cheaper; this tool models chat/completion pricing. How accurate is the projection? Only as good as your average token counts - sample real traffic, since a few long requests can skew the mean. Re-run with measured averages before you commit a budget, and revisit when prices or models change.

Frequently Asked Questions

Related Developer Tools Calculators