LLM Token Cost Calculator — GPT, Claude & Gemini Pricing

Quick Start Guide

  1. Select a Model: Choose the LLM you plan to use (GPT-5.6, Claude, Gemini, etc.).
  2. Choose Input Mode: Enter tokens directly or use word counts — the calculator converts automatically.
  3. Enter Input Tokens: Input the number of tokens (or words) for your prompts.
  4. Enter Output Tokens: Input the expected number of tokens (or words) in the response.
  5. View Cost Breakdown: See the total cost and per-token breakdown instantly.
  6. Compare Models: Switch to Compare mode to see costs ranked across multiple models side by side.

Understanding LLM Token Costs

What is a token?

LLMs bill by the token, not the word. A token is a chunk of text — roughly 4 characters, or about three-quarters of a word in English. Your cost is just the input text plus the generated output text, each charged at the model's per-token rate.

Input vs output pricing

Providers charge a separate rate for input (what you send: the prompt, system message, and any context) and output (what the model writes back). Output is usually the more expensive of the two, so a long answer often costs more than a long prompt.

Estimating monthly cost

Per-request cost is (input tokens × input rate) + (output tokens × output rate). Multiply that by your requests per month for a monthly estimate. Set the requests-per-month field to project recurring spend before you launch.

Cached input and batch pricing

If you resend the same context (a long system prompt or a document) across requests, cached-input pricing can lower the cost of those repeated tokens. Batch APIs trade speed for a discount on non-urgent jobs. Advanced mode lets you apply both.

How to reduce token costs

Trim the prompt and context to what the model actually needs, cap the output length when you don't need long answers, use a smaller model for simple tasks and a larger one only when quality demands it, and reuse cached context and batch non-urgent work where the API supports it.

Common mistakes

Counting only the prompt and forgetting output tokens (often the bigger half); ignoring the system prompt and conversation history, billed on every request; comparing models on price alone without checking quality for your task; and reading rates per 1K when the provider quotes per 1M, or vice-versa.

Features

Multiple Models: GPT-5.6, Claude, Gemini, DeepSeek, Mistral, xAI and more. Compare costs across providers.

Token & Word Modes: Enter tokens or words; calculator converts. Use requests per month to see projected monthly cost.

Cost & Monthly Projection: Per-request cost and optional monthly estimate. Pricing verified against provider API pages.

Scenario Mode: Realistic presets for common workloads like support chat and content generation.

Export Data: Export to JSON, CSV, or PDF for budgeting and reporting.

Common Use Cases

Customer Support Chatbot: average 200 input tokens per conversation and 300 output tokens per response — compare GPT-5.6 Terra against a cheaper tier like GPT-5.6 Luna for cost optimization.

Content Generation at Scale: 1,500-word articles run roughly 2,000 tokens input, 6,000 tokens output — compare Claude Sonnet 5 against GPT-5.6 Terra for quality vs cost, and use custom pricing if you have volume discounts.

Code Generation & Review: code review requests run 500-1,000 input tokens with 200-500 output tokens, and generation runs 300 input with 1,000-3,000 output — a useful GitHub Copilot alternative cost comparison.

Research & Analysis: upload documents (use the Token Converter to estimate input tokens) and generate summaries at roughly 5,000 input tokens, 500 output tokens.

Frequently Asked Questions

Per-request cost = (input tokens × input rate per 1M) + (output tokens × output rate per 1M). Each model has different rates — GPT-5.6 Terra, for example, charges $2.00/$12.00 per 1M tokens. In advanced mode, cached-input and batch discounts are applied before the monthly projection.
It depends on your token volumes and quality needs. Smaller models like GPT-5.6 Luna and Claude Haiku 4.5 are the most affordable. GPT-5.6 Terra and Claude Sonnet 5 sit in the mid-tier. Use Compare mode in this calculator to run the same workload across models and see ranked costs side by side.
Enter your average input and output tokens per request, then set your expected requests per month. The calculator multiplies per-request cost by that volume to give a monthly projection. Adjust cached-input or batch discounts in advanced mode if your usage qualifies.
Scenario mode loads realistic presets for common workloads — support chat, content generation, and others — so you can start with typical token volumes and tweak from there instead of estimating from scratch.
Yes. Compare mode runs the same token workload across selected models and ranks estimated cost. This helps you evaluate cost tradeoffs before switching providers or model tiers.
Input tokens are the text you send (prompts, context, instructions). Output tokens are what the model generates. Output tokens are typically 3-5x more expensive than input tokens. A token is roughly 4 characters or 0.75 words in English.
You most likely crossed the model's long-context threshold. OpenAI, xAI and Google all charge a higher rate for large prompts — GPT-5.6 Terra, for instance, moves from $2.00/$12.00 to $4.00/$18.00 per 1M once input passes 272,000 tokens, and xAI and Gemini switch at 200,000. The catch is that the higher rate applies to the whole request, not just the tokens past the line, so a prompt one token over the threshold costs roughly double one just under it. The calculator applies this automatically and says so in the adjustments list under your result.
Yes. Switch to word mode and the calculator converts to tokens automatically using the standard ratio of ~0.75 words per token. This is useful when you know your document length in words but not in exact token counts.

Last updated