Free tool
AI Token Cost Calculator
Enter your request volume and tokens per request to estimate and compare API costs across models with live EasyAPI pricing, and find the best value for your workload.
Usage
Share of input tokens served from the prompt cache. Only applies to models with a cache price.
Some models price by time of day. Auto uses the model billing timezone to pick the current window.
Estimate tokens from text
A rough estimate: about 1 token per Chinese character and about 4 characters per token for English. Tokenizers differ by model; the usage returned by the API is authoritative.
Cost comparison
deepseek-v4.1-flash
qwen3.8-flash
qwen3.5-flash
glm-5.3-flashqwen3.5-flash is the cheapest at about ¥9.82 per month, saving about 60% compared with deepseek-v4.1-flash.
| Model | Input / 1M tokens | Output / 1M tokens | Per request | Per day | Per month | Per year | Expand details |
|---|---|---|---|---|---|---|---|
qwen3.5-flash AlibabaLowest cost | ¥0.1652 | ¥1.6345 | ¥0.000982 | ¥0.327 | ¥9.82 | ¥118 | |
glm-5.3-flash Z.AI | ¥0.5528 | ¥1.8425 | ¥0.00147 | ¥0.491 | ¥14.74 | ¥177 | |
qwen3.8-flash Qwen | ¥0.9286 | ¥2.7859 | ¥0.00232 | ¥0.774 | ¥23.22 | ¥279 | |
deepseek-v4.1-flash DeepSeekPeak price | ¥0.8188 | ¥3.2751 | ¥0.00246 | ¥0.819 | ¥24.56 | ¥295 |
Prices come from live EasyAPI Model Square pricing (platform discounts included). Tiered models use the tier matching per-request input length; peak/valley models use the selected period. Daily cost assumes 30 days per month; your bill is authoritative.
How costs are calculated
LLM APIs usually bill input and output tokens separately, priced per million tokens. The calculator uses these formulas:
- Cost per request = (non-cached input tokens × input rate + cached input tokens × cache rate + output tokens × output rate) ÷ 1,000,000
- Monthly cost = cost per request × requests per month; daily = monthly ÷ 30; yearly = monthly × 12
- For tiered models (priced by input length), the tier matching your per-request input tokens is applied; expand a row for the full tier table
- For time-of-day models, switch between valley and peak rates; the default follows the model billing timezone
- Output rates are usually several times input rates, so long-reply workloads (code, long-form writing) are output-dominated
- With prompt caching, repeated system prompts and context are billed at the lower cache rate, cutting long-context costs significantly
- A model row is flagged as over context when input + output exceed its context window; such requests actually fail, so lower the token amounts
FAQ
What is a token?
A token is the basic unit an LLM uses to process text. Roughly, one Chinese character is about one token, and in English one token is about 4 characters or 0.75 words. Models bill input and output tokens separately.
How accurate are the prices?
Prices are read live from EasyAPI Model Square, include platform discounts, and match the model detail pages. Tiered models are priced using the tier that matches your per-request input length; peak/valley models use the selected period (the current one by default). Your console bill remains authoritative.
How do I know how many tokens a request uses?
Use “Estimate tokens from text” above for a rough figure. Once integrated, the exact usage of every call appears in the usage field of the API response and in your console usage logs.
How can I reduce LLM API costs?
Common approaches: use cheaper lightweight models for simple tasks, trim prompts and context, cap the maximum output length, enable prompt caching, and route requests across models by task complexity.
Can one API key call every model in the calculator?
Yes. EasyAPI offers an OpenAI-compatible API, so one API key can call any model your account has access to. Switching models only requires changing the model parameter.
Call all these models with one API key
Sign up for EasyAPI to call GPT, Claude, Gemini, DeepSeek and more through an OpenAI-compatible API, paying only for what you use.