AI API Cost Calculator: Integrated Comparison of OpenAI·Claude·Gemini (as of June 2024)

Comparison of Pricing Policies by Major AI API Providers

Input token price: GPT‑4o (as of May 2024) 0.005 USD per 1K tokens, GPT‑4 Turbo 0.01 USD per 1K tokens, GPT‑3.5 Turbo 0.0005 USD per 1K tokens
Output token price: GPT‑4o 0.015 USD per 1K tokens, GPT‑4 Turbo 0.03 USD per 1K tokens, GPT‑3.5 Turbo 0.0015 USD per 1K tokens
Free trial: $5 worth of free credits (usable for 90 days)
Limits: Requests per second (10–30 RPS), daily request limit (10,000–100,000 requests)
Recommended for: Complex natural‑language processing, code generation, multilingual support

Input token price: Claude 3 Opus 0.015 USD per 1K tokens, Sonnet 0.003 USD per 1K tokens, Haiku 0.00025 USD per 1K tokens
Output token price: Claude 3 Opus 0.075 USD per 1K tokens, Sonnet 0.015 USD per 1K tokens, Haiku 0.00125 USD per 1K tokens
Free trial: $5 worth of free credits (usable for 90 days)
Limits: Requests per second (5–20 RPS), daily request limit (5,000–50,000 requests)
Recommended for: Real‑time conversation, document summarization, code review

Input token price: Gemini 1.5 Pro 0.0075 USD per 1K tokens, Flash 0.000125 USD per 1K tokens
Output token price: Gemini 1.5 Pro 0.03 USD per 1K tokens, Flash 0.000375 USD per 1K tokens
Free trial: $3 worth of free credits (usable for 90 days)
Limits: Requests per second (10–50 RPS), daily request limit (10,000–100,000 requests)
Recommended for: Multimodal (text + image) processing, large‑scale document analysis

Count tokens: Add the tokens in the request text and the response text. (Example: 1,000 words ≈ 1,500 tokens)
Apply prices: Multiply the number of input and output tokens by the respective model prices.
- Example 1) Using GPT‑4o for a 500‑word request (input) and a 1,000‑word response (output)
  - Input: 750 tokens × 0.005 USD = 0.00375 USD
  - Output: 1,500 tokens × 0.015 USD = 0.0225 USD
  - Total: 0.02625 USD (≈ 35 KRW)
- Example 2) Using Gemini 1.5 Flash to analyze a 10,000‑word document (input only)
  - Input: 15,000 tokens × 0.000125 USD = 0.001875 USD
  - Output: 0 (only analysis result requested)
  - Total: 0.001875 USD (≈ 2.5 KRW)
Estimate monthly cost: Multiply daily request count by 30 days and by the average cost per request.

Model selection: Use GPT‑4o for complex tasks, GPT‑3.5 Turbo or Gemini Flash for simple tasks
Caching: Store identical responses locally to avoid repeated calls
Token optimization: Simplify prompts and minimize unnecessary output
Leverage free tier: Test within the free‑credit period before switching to paid usage

Feature	OpenAI (GPT‑4o)	Claude (Sonnet)	Google (Gemini 1.5 Pro)
Code generation	Excellent (Python, JS, etc.)	Good (Python, Go)	Fair (Python, C++)
Document summarization	Good (long‑form)	Best (structured)	Best (multimodal)
Real‑time conversation	Fair	Best (fast response)	Fair
Image processing	Limited	Limited	Best (Gemini Vision)
Multilingual support	80+ languages	100+ languages	100+ languages

Request rate (RPS): Claude Haiku 5 < OpenAI GPT‑3.5 10 < Gemini Flash 50
Daily request limit: Claude Opus 5,000 < OpenAI GPT‑4 10,000 < Gemini Pro 100,000
Latency: Claude 3 Opus 1–2 s < GPT‑4o 1–3 s < Gemini 1.5 Pro 2–5 s

Developers / code generation: OpenAI GPT‑4o (high code quality) or Claude Sonnet (fast feedback)
Enterprise document analysis: Google Gemini 1.5 Pro (multimodal, large‑scale processing)
Customer‑service chatbots: Claude Haiku (low cost, fast response)
Multilingual services: Claude 3 (100+ languages) or Gemini (integrated with Google ecosystem)
Budget‑conscious projects: Gemini Flash (lowest input token price) or GPT‑3.5 Turbo (cheap output tokens)