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.
Paste some text and this section will list each model's token estimate and single-input cost.
Per-model token estimate and single-input cost
| Model | Chinese factor | Estimated tokens | Cost per call (USD) | Cost per call (CNY) |
|---|
Prices and factors updated 2026-07; refer to each vendor's official site. Chinese is counted per character and English per word (uniformly ≈1.3 tokens/word); cost assumes the entire text is used as input.
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.
Enter the input/output tokens per call and the number of calls per month (calls must be greater than 0); the result calculates automatically.
Monthly cost comparison
| Model | Price (input / output, $/million) | Cost per call (USD) | Monthly cost (USD) | Monthly cost (CNY) |
|---|
Prices updated 2026-07; refer to each vendor's official site. Cache discounts, batch discounts, and free tiers are not included — this is a budget-level upper-bound estimate.
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 →