如何计算AI API Cost
learn.whatIsHeading
An AI API cost calculator estimates the cost of using large language model APIs based on token usage. AI APIs charge separately for input (prompt) and output (completion) tokens.
公式
cost = (input_tokens × rate_in) + (output_tokens × rate_out)
- T_in
- Input tokens (tokens) — Tokens sent to API
- T_out
- Output tokens (tokens) — Tokens generated by API
- rate_in
- Input rate ($/1M tokens) — Price per million input tokens
- rate_out
- Output rate ($/1M tokens) — Price per million output tokens
分步指南
- 1Cost = (input tokens ÷ 1000) × input rate + (output tokens ÷ 1000) × output rate
- 2Rates are per 1,000 tokens (or per million tokens on newer models)
- 3Monthly cost = cost per request × monthly request volume
- 4Input tokens are typically cheaper than output tokens
例题解析
输入
GPT-4o: 500 input + 200 output tokens, 1000 req/month
结果
~$3.25/month
输入
Claude Haiku: same usage
结果
~$1.20/month
输入
GPT-4o-mini: 1M requests/month
结果
~$150–300/month typical
常见问题
Why is output more expensive than input?
Generating tokens is computationally more costly than reading them. Output pricing reflects the higher compute cost.
How do I estimate total API costs?
Calculate (input tokens ÷ 1,000,000) × rate_in + (output tokens ÷ 1,000,000) × rate_out. Track all API calls.
What are typical LLM token rates?
As of 2026, rates vary: Claude 3.5 Sonnet ~$3/$15 per 1M tokens; GPT-4 ~$30/$60; open-source may be cheaper.
准备好计算了吗？尝试免费的 AI API Cost 计算器
自己尝试一下 →