Cost per request
Combine average input and output token use with provider rates so two models can be compared on the same workload unit.
FREE AI FINOPS TOOL
LLM Cost Calculator
Estimate cost per request, monthly model-provider spend, annual AI cost and the hidden cost of retries or fallbacks. Enter the current token prices from the model provider you actually use.
Editable
token prices
Live
cost model
Free
no signup
Live calculator
All prices are editable. Use the current token rates from your provider.
WORKLOAD INPUTS
PROVIDER TOKEN PRICING
Provider prices change over time and can differ by model, cache state, region or modality. This calculator intentionally does not hard-code vendor pricing.
ESTIMATED MODEL-PROVIDER SPEND
Monthly estimate
$824
103K effective model calls including overhead
Cost / request
$0.0080
Cost / 1K requests
$8.00
Annual estimate
$9,888
Retry/fallback overhead
$24.00
Monthly token volume
175.1M tokens
Input cost 37.5%Output cost 62.5%Base monthly $800
Estimate only. It excludes platform subscriptions, non-token modalities, cached-token rules, batch discounts, taxes and provider-specific pricing behavior unless you incorporate those effects into the rates you enter.
HOW TO USE THE NUMBER
Move from token price to workload economics.
Provider price per million tokens is only the starting point. Production teams need to understand what one request, one agent run and one month of traffic actually cost.
Reliability overhead
Retries and fallbacks create additional inference. Model that overhead explicitly instead of treating every user action as one provider call.
Monthly unit economics
Multiply workload economics by real request volume, then compare the result with budgets, product revenue or operational value.
FORMULA
What the calculator is estimating.
Base cost per request = (input tokens ÷ 1,000,000 × input price) + (output tokens ÷ 1,000,000 × output price).
Monthly model spend multiplies that request unit by monthly request volume. Retry/fallback overhead increases the effective number of model calls, which is useful for modeling reliability behavior that would otherwise disappear from a simple token estimate.
For production decisions, compare this estimate with actual request-level telemetry. Cached tokens, batch pricing, images, audio, tool calls and provider-specific billing rules can change the final invoice.
FAQ
LLM cost calculator questions.
How does the LLM cost calculator work?
It multiplies average input and output tokens per request by the token prices you enter, then applies monthly request volume and optional retry or fallback overhead. The result includes cost per request, monthly model-provider spend and an annual estimate.
Does the calculator use current OpenAI, Anthropic or Gemini pricing automatically?
No. Provider pricing changes frequently and can vary by model, cache state, modality, region or billing program. The calculator intentionally keeps input and output prices editable so you can enter the current rates from the provider you are evaluating.
Does this estimate include Clyvel pricing?
No. The calculator estimates model-provider inference spend from the token rates and workload assumptions you enter. Platform subscriptions, taxes and other provider-specific billing rules are separate.
How should I estimate cost for an AI agent?
Use the average total model calls and tokens generated by one logical agent run, or calculate each model-call type separately and add the results. Agent retries and fallback attempts can be represented with the overhead percentage.