Cost estimate
Inference cost calculator.
Choose a model and enter the input and output tokens you expect. The calculator compares their cost at the model's reference list price with their cost through Avenro, at the prices the API bills with. No account or API key is needed, and nothing you enter leaves your browser.
claude-fable-5-1 · Anthropic
50,000,000 tokens
10,000,000 tokens
| Per 1M tokens | Reference | Avenro |
|---|---|---|
| Input | $10.00 | $6.00 |
| Output | $50.00 | $30.00 |
- Reference: Anthropic API list price for claude-fable-5-1, checked Oct 7, 2026. Source (opens the provider's price list)
- Reference cost
- $1,000.00At the reference list price
- Avenro cost
- $600.00At Avenro's price
- Difference
- $400.00Avenro costs less
- Savings
- 40%Of the reference cost
Estimates are based on your selected model and entered input/output token volumes using the current configured pricing. Actual costs may vary with real usage and applicable billing rules.
The reference cost uses the provider's public list price as configured for the model; what a provider charges your own account can differ.
Prices from Avenro's catalog as of Oct 14, 2026, 2:52 PM UTC.
How the estimate works
- Prices
- The Avenro price is what your credits are charged per million tokens, the same prices the API bills with. The reference price is the provider's public list price for the same model; the calculator names its source and the date it was last checked. Both come from Avenro's model catalog, and the time they were read is shown with the results.
- The arithmetic
- Reference cost = (input tokens × reference input price + output tokens × reference output price) ÷ 1,000,000, and the Avenro cost the same way with Avenro's prices. Difference = reference cost − Avenro cost, below zero when Avenro would cost more. Savings = difference ÷ reference cost × 100; with a reference cost of $0 there is no percentage.
- Rounding
- Everything is worked out exactly, and only the figures shown are rounded: dollars to the cent (amounts under a cent to six decimals, anything under $0.000001 as <$0.000001), and the saving down to a tenth of a percent, so it is never overstated. With a mouse, point at a dollar figure to see every digit.
- What it leaves out
- Every input token is priced at the input price, without cached-input prices. GPT-6 Astra and Grok 4.6 bill a whole request at higher prices once its prompt passes a set length (shown with the model); the estimate uses the base prices. Billing also rounds each request up to the next $0.000000001. See credits and billing.
- Your numbers stay with you
- The calculator runs in your browser: the token counts you enter are never sent to Avenro or stored, and no model is called.
- Reference prices
- A reference price is the provider's public list price as Avenro's catalog records it for that model. What a provider charges your own account can differ (other tiers, discounts or commitments), so the comparison is with list prices.