Free Online Tool

Chinese Token Calculator

Paste text or enter your call volume to estimate token counts for mainstream LLMs and compare API costs (in RMB).

Text estimate: token count and single-input cost

Paste your text to count Chinese characters and English words in real time, convert them to tokens using each model's estimation factor, and price this text as a single input at each model's input rate. Everything is processed locally in your browser and never uploaded.

Your draft is saved locally and survives a page refresh.

Both tables use this rate for CNY conversion; feel free to change it.

0Chinese characters
0English words

Paste some text and this section will list each model's token estimate and single-input cost.

Usage cost comparison: monthly cost by model

Enter the input/output tokens per call and the number of calls per month to compare each model's monthly bill. Not sure about per-call usage? Paste a typical input above first and use the estimated token count as a reference.

Estimation method and data sources

  • Token estimation factors: Chinese is counted per character — the Claude family ≈1.5 tokens/char, GPT (o200k tokenizer) ≈1.2 tokens/char, DeepSeek ≈0.6 tokens/char (per official docs); English is uniformly ≈1.3 tokens/word. Digits, punctuation, and the like are not converted separately.
  • This is a budget-level estimate: each vendor's tokenizer is implemented differently, so the real token count for the same text will vary. Claude's exact value can be measured with the official count_tokens API; OpenAI's can be verified with its official tokenizer tool.
  • Claude Sonnet 5 launch pricing: before 2026-08-31 it is billed at launch pricing (input $2 / output $10 per million tokens); the table uses the standard price of $3 / $15, so your actual bill before that date will be lower (marked with * in the table).
  • Prices and factors updated 2026-07, all standard API prices (USD per million tokens); refer to each vendor's official site. The cost-accounting approach comes from the tool-cost methodology in the main site's "Complete Guide to AI Content Workflows."
  • No data is uploaded: this page is a pure front-end tool. Pasted content is computed locally in your browser and is never sent to any server.

Want to bring this approach to your own brand and business? Let's talk →