13 tools
LLM Cost Calculators
Turn token counts into the number that appears on an invoice.
A rate per million tokens tells you almost nothing on its own. What matters is your request volume, your input-to-output ratio, how much of your prompt is cacheable, and whether your workload compounds — conversations and agent loops both resend their history, so cost grows faster than request count.
Every price used here comes from the provider’s own documentation and carries the date it was read. A calculator built on a stale rate is confidently wrong, which is worse than having no calculator.
- LLM API Cost CalculatorModel your monthly API bill before you commit to a model.
- Chatbot Conversation Cost CalculatorMulti-turn chat costs far more than turns × price. See by how much.
- LLM Model FinderFilter every model by budget, context and provider in one pass.
- API Rate Limit CalculatorFind out whether your throughput fits inside your tokens-per-minute quota.
- Agent Loop Cost SimulatorWatch what an agent costs iteration by iteration, including the history it drags.
- Subagent Token Budget PlannerSplit a token budget across subagents, with the overhead counted honestly.
- Tokens per Dollar CalculatorHow much text a budget actually buys, in pages and novels.
- Batch API Savings CalculatorHalf price for work nobody is waiting on.
- Fine-Tuning Cost and Payback EstimatorTraining cost, inference premium, and how long until it pays back.
- AI Cost per User CalculatorWhat each user costs you, including the heavy tail that breaks margins.
- Reasoning Token Cost EstimatorYou are billed for thinking you never see.
- Model Routing Savings CalculatorSend the easy 70% to a cheap model and see what it saves.
- Annual LLM Budget PlannerMultiple workloads, compounding growth, twelve months.
Which one you need
If you are sizing a feature that does not exist yet, start with the basic cost calculator: volume, prompt size, answer length, and a price. It gives you the order of magnitude, which is usually the decision.
If the feature is a conversation, use the conversation calculator instead. Chat cost does not grow linearly — every turn resends the history, so the twentieth message pays for the nineteen before it, and a linear estimate will be badly wrong by the end of the session.
If it is an agent, the loop calculator is the one that matters, because the number to budget against is worst case: cost per iteration multiplied by the maximum iteration count. That is what appears on the bill during an incident.
Once something is already running and the question is how to spend less, the caching, batching and routing tools are ordered by saving per hour of work. Caching first — it is a configuration flag. Routing last — it takes a week and saves the most.
Questions
Are the prices here current?
Every rate comes from the provider’s own documentation with the date it was read attached, and each model page links to the source. Prices fall without notice, so follow the source link before committing to a volume decision — a calculator built on a stale rate is confidently wrong.
Why is my real bill higher than the estimate?
Three usual reasons. The estimate used the bare user message rather than the assembled prompt with system instructions, retrieved context and tool schemas. Retries were not counted. Or the model is a reasoning model billing for tokens that never appear in the output.
Guides that use these tools
- How to reduce LLM API costs — Nine levers, ordered by what they return per hour of work. Most teams find 40% in the first three.
- Prompt caching, and what it actually saves — The single largest lever on a repetitive workload — and the cases where it does nothing.
- Why chatbot costs grow faster than your user count — Turn 20 pays for turns 1 to 19 again. Why per-request budgets are wrong by 3–4×.
- What an AI agent actually costs to run — Tool schemas and iteration limits multiply. Where agent bills come from, itemised.