如何计算Token Cost

learn.whatIsHeading

The Token Cost Calculator converts text into tokens and calculates the exact API cost for any LLM provider. It helps developers understand the relationship between text length, token count, and cost across different models and pricing tiers.

公式

Cost = (Token Count / 1,000,000) × Price per Million Tokens

W: Word Count (words) — Number of words in the text
T: Token Count (tokens) — Number of tokens after tokenization
P: Price per Million ($/1M tokens) — Model-specific per-million-token rate
C: Cost ($) — Total cost for the given token count

分步指南

1Paste your text or enter a word/character count to estimate tokens
2Select the LLM provider and model for pricing
3Specify whether the text is input (prompt) or output (completion)
4View the token count, cost for this text, and cost per word

例题解析

输入

1,000 words of English text as input to GPT-4o

结果

Approximately 1,333 tokens. Cost: 1,333 / 1,000,000 × $2.50 = $0.0033 (one-third of a cent).

输入

10,000 words of output from Claude 3.5 Sonnet

结果

Approximately 13,333 tokens. Cost: 13,333 / 1,000,000 × $15.00 = $0.20 (twenty cents).

常见错误注意事项

✕Assuming 1 word = 1 token — in English, 1 token averages about 0.75 words or 4 characters
✕Forgetting that code, non-English text, and special characters often tokenize into more tokens per word
✕Not separating input and output costs when budgeting — they have very different rates

常见问题

How many tokens is a typical ChatGPT conversation?

A typical conversation turn is 50-200 tokens for the user message and 100-500 tokens for the response. A 10-turn conversation accumulates approximately 2,000-5,000 total tokens. System prompts add 100-2,000 additional tokens per call depending on complexity.

Do different languages use different numbers of tokens?

Yes, significantly. English is most efficiently tokenized (~1.3 tokens per word). Chinese, Japanese, and Korean use 2-3 tokens per character. Arabic, Hindi, and other scripts may use 3-5 tokens per word. This means non-English API calls can be 2-4x more expensive per equivalent meaning.

准备好计算了吗？尝试免费的 Token Cost 计算器

自己尝试一下 →