<aside> 🎯 Tokens are the fundamental unit of AI API pricing. Understanding them can save you hundreds of dollars per month.

</aside>

What is a Token?

A token is a chunk of text that AI models process. It's not exactly a word — it's more like a piece of a word. On average, 1 token ≈ 0.75 English words, or about 4 characters.

Token Examples

• "Hello" = 1 token
• "Artificial Intelligence" = 2 tokens
• "supercalifragilisticexpialidocious" = 7 tokens
• "你好世界" (Chinese) = 4 tokens (CJK characters use more tokens)

Why Tokens Matter for Developers

Every AI API charges by tokens. When you send a prompt and receive a response, you're paying for:

1. Input tokens — your prompt (system message + user message + context)
2. Output tokens — the model's response
3. Total cost = (input tokens × input price) + (output tokens × output price)

Token Pricing Comparison (2026)

• GPT-4o: $2.50 / 1M input, $10.00 / 1M output (direct from OpenAI)
• GPT-4o via Crazyrouter: ~$1.12 / 1M input, ~$4.50 / 1M output (55% cheaper)
• Claude Sonnet 4: $3.00 / 1M input, $15.00 / 1M output
• Claude Sonnet 4 via Crazyrouter: ~$1.35 / 1M input, ~$6.75 / 1M output
• DeepSeek R1: $0.55 / 1M input, $2.19 / 1M output