AI API Cost Calculator

Estimate AI API cost from tokens, requests, model pricing, traffic, caching, and usage scenarios.

Monthly API cost

$604.22

30 active days

Daily API cost

$20.14

Input plus output token cost

Cost per request

$0.00081

Average blended request cost

Monthly tokens

840M

Input and output combined

Token pricing inputs

Enter editable pricing per 1M tokens and usage assumptions.

Daily input cost

$9.14

18.3M billable input tokens

Daily output cost

$11.00

5.5M output tokens

Raw input tokens

22.5M

Before cache discount

Annual run rate

$7,250.63

Monthly estimate multiplied by 12

Related calculators

What is the AI API cost calculator?

This calculator estimates AI API cost from input tokens, output tokens, daily request volume, model price, and exchange rate. It supports model comparison across OpenAI, Anthropic, Google, xAI, and DeepSeek style options in the Korean source page.

Tokenization matters. English averages roughly 4 characters per token, while Korean can use about 2 to 3 tokens per Korean character in some tokenizers. For example, the Korean greeting “annyeonghaseyo” has 5 Korean characters and can become about 10 to 15 tokens.

Cost formula

Request cost = input tokens x input price / 1,000,000 + output tokens x output price / 1,000,000. Monthly cost = request cost x daily requests x 30. Output tokens are often 2 to 5 times more expensive than input tokens.

Source price examples

February 2026 model price table in the source

  • GPT-5.2 input USD 1.75 and output USD 14.00 per 1M tokens.
  • Claude Opus 4.6 input USD 5.00 and output USD 25.00 per 1M tokens.
  • Claude Sonnet 4.5 input USD 3.00 and output USD 15.00 per 1M tokens.
  • Gemini 2.5 Pro input USD 1.25 and output USD 10.00 per 1M tokens.
  • Grok 4 input USD 3.00 and output USD 15.00 per 1M tokens.
  • DeepSeek R1 input USD 0.55 and output USD 2.19 per 1M tokens.
  • GPT-5 nano input USD 0.05 and output USD 0.40 per 1M tokens; Gemini 2.5 Flash input USD 0.15 and output USD 0.60 per 1M tokens.

GPT-4o style request example

With 1,000 input tokens, 500 output tokens, input price USD 2.50 per 1M, and output price USD 10.00 per 1M, the cost is USD 0.0075 per request.

Monthly scenarios and optimization

Customer-support chatbot example

  • For 500 requests per day, average 2,000 input tokens and 1,000 output tokens, GPT-5.2 costs about USD 262.5 per month, or about KRW 381,000 in the source exchange-rate example.
  • The same workload on Claude Sonnet 4.5 costs about USD 315 per month, or about KRW 457,000.
  • The same workload on GPT-5 nano costs about USD 7.5 per month, or about KRW 11,000.
  • A simple chatbot can cut cost by more than 90% with a lightweight model, while complex premium tasks may still need a stronger model.

Cost reduction levers

  • Use tiered model routing: simple FAQ to GPT-5 nano or Gemini Flash, complex premium cases to larger models. Moving 70% of simple Q and A traffic to a mini model can reduce cost by more than 60%.
  • Prompt engineering can save 20% to 40% of tokens by removing unnecessary examples, repetition, and verbose formatting.
  • OpenAI Batch API can reduce cost by 50% for non-real-time work.
  • Anthropic Prompt Caching can save up to 90% for repeated system prompts, and application-side caching avoids paying again for repeated answers.

API prices change frequently. This English page preserves the source document’s February 2026 assumptions; production planning should verify current provider pricing before committing spend.