LLM API cost
LLM API cost calculator
Monthly and per-request cost of an LLM API workload from input tokens, output tokens, requests a month, cached share and the Batch discount.
Prices read from the vendor pages on 2026-10-20. Arithmetic on list prices, not advice or a quote.
Saving and exports need Norwell Tools Pro, $9 a month. See what is in it. Everything above is free.
How it works
Vendors price input and output tokens separately per million tokens. Output tokens cost several times more than input, so a chatty assistant costs more than a long-prompt classifier with the same total token count. Cached input tokens are billed at the cache-read price (often a tenth of the input price or less) and the Batch API halves both prices at most vendors in exchange for asynchronous delivery.
The formula
cost_per_request = (input_tokens * ((1 - cached) * input_price + cached * cached_price) + output_tokens * output_price) / 1,000,000 * (batch ? 1 - discount : 1); monthly = cost_per_request * requestsWorked example
With the inputs the page opens on:
- model: claude-sonnet-5-5
- input tokens: 1500 tokens
- output tokens: 500 tokens
- requests per month: 100000
- cached share pct: 0%
- batch: false
gives
- cost per request: $0.00800
- cost per 1k requests: $8
- monthly cost: $800
- monthly input cost: $300
- monthly output cost: $500
- input share pct: 37.5 %
Sources
- https://developers.openai.com/api/docs/pricing (read 2026-10-20)
- https://platform.claude.com/docs/en/about-claude/pricing (read 2026-10-20)
- https://ai.google.dev/gemini-api/docs/pricing (read 2026-10-20)
- https://docs.mistral.ai/inference/pricing (read 2026-10-20)
- https://api-docs.deepseek.com/quick_start/pricing (read 2026-10-20)
- https://docs.x.ai/docs/models (read 2026-10-20)
Prices read from each vendor's own pricing page on 2026-10-20; the page may have changed since. This is arithmetic on published list prices, not advice or a quote. Check the vendor's page before committing a budget.
Questions
Where do the prices come from?
Does this include thinking or reasoning tokens?
Why does my bill differ?
Related tools
The same workload priced on every model in the table, cheapest first.
What one prompt costs on a model from its length in characters and the expected answer length.
How many requests and tokens a monthly budget buys on a model.
Convert a token price between per 1M, per 1K and per token, and price any token count.