LLM token cost calculator
Enter what you send and generate in a month. Every model in our price table is costed the same way, sorted cheapest first — including the ones that are not ours.
The presets sit at a 4:1 input-to-output ratio, which is roughly where chat and agent traffic runs. Change either field to use your own numbers.
| Model | Input cost | Output cost | Total / month | vs cheapest on any plan |
|---|
Pay-as-you-go rates, before any plan allowance. Prices verified August 14, 2026 — sources on /pricing. Claude and GPT figures come from the vendors’ own pricing pages.
What this leaves out
- Reasoning tokens. Billed as output, invisible in your prompt and your reply. On a reasoning model, treat the output number as a floor.
- Retries. An agent that loops twice costs twice, whatever the rate.
- Cached input. Some providers discount repeated prefixes heavily. If your prompts share a long prefix, that can matter more than the headline rate.
- Quality. The cheapest row that fails your task is not the cheapest option. Run your own evals before you move traffic.
FAQ
▸How do I estimate how many tokens I use?
A token is roughly 0.75 English words, so 1,000 words is about 1,300 tokens. Code and non-English text tokenise less efficiently. If you already run the workload, take the token counts from your provider’s usage page — an estimate from a real week beats any calculator.
▸Why does the input/output split matter so much?
Output usually costs three to six times more per token than input. Two workloads with the same total token count can differ several-fold in price depending on which side of that split they sit on. Chat and agent traffic is heavily input-weighted; content generation is not.
▸Does this include reasoning tokens?
No, and neither does any calculator that only takes two numbers. Models that think before answering bill those thinking tokens as output even though you never see them. If you use a reasoning model, treat the output figure here as a floor.
▸Is a subscription cheaper than this?
It can be. A plan holds a fixed monthly allowance that buys tokens at these same rates, sized so it is worth several times its price — the saving comes from actually spending it. If your usage is spiky or small, pay-as-you-go is cheaper. Compare your monthly figure here against the plan prices.