What does your LLM workload actually cost per month?
Plug in your monthly request volume and average token sizes. The calculator prices the same workload three ways — Router One's list price before discount, Router One pay-as-you-go after discount, and Router One subscription plans — using live rates from the model catalog. Every figure on this page is Router One's own posted rate; no provider's price list is quoted or used.
Start from a profile
Monthly volume: 36.0M input + 7.2M output tokens
Assumes every request has the entered token sizes; excludes prompt caching. Estimate groups in different input-length tiers separately.
Router One list price (before discount)
$540/mo
Router One's own posted rate for this model, before any current discount.
Router One pay-as-you-go
$162/mo
−70% vs Router One list price
Discounted rate on this model, below Router One's list price.
Router One subscription
$20.19/mo
Router.One Pro · quota covers this request volume
- Router.One Pro · $20.19/mo6,500 calls / mo
- Router.One Max · $47.99/mo13,650 calls / mo
- Router.One Ultra · $149/mo45,450 calls / mo
Model-tier quotas apply — see current quotas on /pricing. /pricing
How the estimate is computed
- Token math: monthly cost = requests × (input tokens × input rate + output tokens × output rate) ÷ 1,000,000.
- For models with long-context tiers, the entered input tokens per request select the rate for the whole request, including output. The estimate treats every request as the same size. If real requests fall into different tiers, estimate each group separately and add the results; monthly token totals do not select a tier.
- Rates are pulled live from the public model catalog (/models) — the same posted per-token prices your requests are billed at.
- Both price columns come from that one catalog: the first is the selected model's undiscounted list rate, the second is its discounted rate where a discount currently applies.
- Prompt caching is not modeled. Whether caching applies at all is a property of the model API you call and of how the request is shaped, so the estimate assumes none of it — check your own per-request traces for what actually happened.
- The subscription column checks your request volume against each plan's monthly call quotas; model-tier quotas apply, see /pricing for details.
The three columns, honestly labeled
Router One list price (before discount)
What the workload costs at Router One's own posted rate for the selected model, before any current discount is applied. This is our undiscounted catalog rate — not a provider's published price list, and not a per-model official-price comparison.
Router One pay-as-you-go
The same workload at the rate you are actually billed: the discounted rate where one currently applies, otherwise the list rate above. Router One's discounts run up to 90% off official prices on select models, and this column shows the resulting Router One price.
Router One subscriptions
Flat monthly plans with metered call quotas by model tier. When your volume fits a plan's quota, the flat price replaces the per-token bill entirely; overage bills to wallet balance at posted rates, so nothing cuts off mid-session.
FAQ
Where do the prices in the calculator come from?
Live from the public model catalog and subscription plan APIs — the same data shown on /models and /pricing. Nothing is hardcoded, so the calculator stays in sync with actual billing rates.
Is the first column another provider's price?
No. Both price columns are Router One's own catalog rate for the model you picked — before discount and after discount. No provider's published price list is quoted here. For a feature-by-feature gateway comparison, see /openrouter-alternative.
Why do the two Router One columns sometimes show the same number?
Discounts apply to select models, up to 90% off official prices. On a model with no current discount, pay-as-you-go is the list rate, so both columns match — for heavy use on those models the subscription plans are usually the bigger lever.
Are subscription quotas really enough for heavy Claude Code use?
Whether a subscription quota covers heavy Claude Code use depends on your monthly call volume — that is exactly what the calculator checks. Quotas are metered per month by model tier, and overage bills to your wallet balance at posted rates instead of cutting you off. Current quotas are on /pricing.
Is this estimate exact?
It's an estimate. Every request is treated as having the entered input and output sizes, and the input size selects its long-context price tier. If requests cross different thresholds, estimate each group separately; an average can hide a higher-priced group. Prompt caching is not modeled. Your dashboard shows the final per-request cost alongside model, tokens, latency, and status.