Question 1

How is LLM API cost calculated?

Accepted Answer

Providers bill input (prompt) tokens and output (completion, including reasoning/thinking) tokens separately, priced per 1 million tokens. Cost per request = input tokens × input price / 1,000,000 + output tokens × output price / 1,000,000. Cached input tokens are billed at the lower cached-input rate instead of the normal input rate.

Question 2

How many tokens is a word or a character?

Accepted Answer

For typical English text, 1 token is roughly 4 characters or about 0.75 words, so 1,000 words is about 1,333 tokens. Code, non-English languages and unusual formatting often use more tokens. This calculator uses those rules of thumb when you enter words or characters; use the provider's tokenizer or the usage field in API responses for exact counts.

Question 3

What is cached input and how much does it save?

Accepted Answer

Prompt caching lets a provider reuse a repeated prompt prefix (system prompt, documents, tool definitions, conversation history). Cache hits are billed at a fraction of the normal input price, often 2.5% to 10% depending on the model. Writing to the cache can cost more than normal input on some providers (for example Anthropic's 5-minute and 1-hour cache writes), which this calculator does not include.

Question 4

Why are output tokens more expensive than input tokens?

Accepted Answer

Generating each output token requires a full sequential pass of the model, while input tokens are processed in parallel. Output usually costs 4 to 8 times more than input. For reasoning models, hidden thinking tokens are billed as output, so actual output can be much larger than the visible answer.

Question 5

How much does Claude Code or another coding agent cost per month on the API?

Accepted Answer

It depends heavily on usage. Coding agents resend large contexts (files, tool results, history) on every step, so input dominates and prompt caching matters a lot. The preset here assumes 150 agent steps per day, 50,000 input tokens per step with 85% cache hits and 1,200 output tokens per step; it is only an estimate. Subscription plans may be cheaper for heavy individual use; check the provider's plan pricing.

Question 6

Are these prices up to date?

Accepted Answer

Prices were copied from each provider's official pricing page and last checked on 2026-10-24. They are standard pay-as-you-go list prices, excluding batch discounts, regional uplifts, long-context surcharges, taxes, tools (web search, code execution) and negotiated discounts. Always confirm on the official page before budgeting.

LLM API Cost Calculator

Frequently asked questions