Question 1

How are AI API token costs calculated?

Accepted Answer

Token costs are calculated by multiplying input token count by input price per million tokens, plus output token count by output price per million tokens. For example, GPT-4 Turbo costs $10 per million input tokens and $30 per million output tokens. A request with 1,000 input tokens and 500 output tokens costs $0.01 + $0.015 = $0.025.

Question 2

What is the difference between input and output token pricing?

Accepted Answer

Input tokens (prompt) are typically cheaper than output tokens (generated response). GPT-4 Turbo: input $10/1M, output $30/1M. Claude 3.5 Sonnet: input $3/1M, output $15/1M. Output tokens require more computation. The ratio varies by model, usually 2:1 to 5:1 for premium models.

Question 3

How many tokens does a typical AI request use?

Accepted Answer

Simple chat messages use 50-200 tokens. Document analysis uses 500-5,000 input tokens. Code generation typically uses 1,000-3,000 output tokens. RAG queries with context use 2,000-10,000 input tokens. Agent workflows can use 10,000-50,000 tokens per session across multiple calls.

Question 4

How can I reduce my AI API costs?

Accepted Answer

Key strategies: use smaller models for simple tasks, cache frequent prompts, limit output token length, implement prompt compression, use batching for multiple requests, choose region-optimized pricing, monitor usage with budgets, and route complex queries to premium models only when necessary.

Question 5

What is context caching and how does it save costs?

Accepted Answer

Context caching stores frequently used prompts (like system instructions or large documents) server-side, reducing input token costs on subsequent requests. Claude offers 90% discount on cached tokens. Gemini offers free caching for 1 hour. Useful for RAG, multi-turn conversations, and repeated system prompts.

AI Token Cost Calculator

Need a production-ready AI cost plan?