MundoKit

Long-context pricing calculator

Enter prompt sizes and request counts to compare the cost at each billing tier. Choose a dated API preset or custom prices; the table also shows the exact threshold boundary.

Inputs
To edit prices, choose Custom. Its initial values match Haiku 5.5; selecting a preset does not copy values into custom fields.
tokens
Include billed reasoning tokens. Same output count for every row.
One row per cohort: input tokens, number of requests. Use plain digits without thousands separators. Maximum 200 rows and 20,000 characters. Include schemas, history and request formatting in input counts.

Results

Distribution total (USD)
$31.700000
Weighted cost per request
$0.031700
Difference from all-standard-rate baseline
$21.200000
Cost increase: threshold to threshold + 1
$0.042000
Boundary comparison and actual prompt distribution
ScenarioInput tokensRequestsRate tierUSD / requestCohort USD
Boundary comparison99,9991Standard$0.010500—
Boundary comparison100,0001Standard$0.010500—
Boundary comparison100,0011Higher$0.052501—
Distribution99,000500Standard$0.010400$5.200000
Distribution101,000500Higher$0.053000$26.500000
  • Higher rates apply to the whole request, not only excess tokens. Output does not trigger the threshold.
  • Uncached token charges only: no cache writes, discounts, tools, taxes or regional premiums. This is a billing comparison, not a context-fit check.
  • Presets checked 2026-11-14; review due 2026-12-14. Custom prices are your assumptions.

Results update as you type. The share link stores your inputs after the # in the address, which is not sent to our server, but anyone you send it to can read them.

Prices change often. We list the date we last checked each price; confirm on the provider’s official pricing page before you budget.

How it works

Each row represents requests of one input size. Compare that size with the threshold, then apply the selected tier to every input and output token in the request.

request cost = (input tokens × input price + output tokens × output price) ÷ 1,000,000
total = sum(request cost × requests)
weighted average = total ÷ total requests

The Claude pricing page sets Haiku 5.5’s threshold at 100,000 input tokens: equality keeps the lower tier. Standard input/output prices are $0.10/$0.50 per million, rising to $0.50/$2.50 above the threshold. GPT-6.1 Sol’s model page sets its threshold at 272,000: $2/$10 becomes $4/$15 above it. These are standard global API prices checked November 14, 2026.

The baseline applies lower prices to the same distribution. It is a comparison, not an available discount. Boundary rows are illustrative and are not added to the distribution total. Rounding: no intermediate rounding; USD results use six decimal places.

Worked examples

One token crosses the Haiku 5.5 boundary

With 1,000 output tokens, 100,000 input tokens cost $0.010500. At 100,001 input tokens the cost is $0.0525005, displayed as $0.052501. The whole request uses the higher tier.

A mixed distribution instead of an average prompt

At Haiku 5.5 rates with 1,000 output tokens, 500 requests of 99,000 input tokens cost $5.20; 500 requests of 101,000 cost $26.50. Total: $31.70, or $0.0317 per request. Applying lower rates to every row would give $10.50.

Frequently asked questions

Is only the part above the threshold charged more?
No. These presets apply the higher input and output rates to the entire request once input exceeds the threshold.
Does generated output trigger long-context pricing?
No. The selected presets use prompt length for the threshold. Output is then billed at the tier selected by that input length.
Can I include prompt caching?
This version calculates uncached requests. Claude counts cache reads and writes toward prompt length, but their prices differ. Use the prompt caching calculator for cache cost scenarios.
How do I edit a preset’s prices?
Choose Custom and enter the threshold and four rates. Custom fields initially contain the Haiku 5.5 values; switching presets does not overwrite your custom fields.
Does a low price mean the prompt fits the model?
No. This calculator evaluates billing only. Use the context window calculator to check input plus reserved output against model capacity.

Sources

Last reviewed 2026-11-14; next review due by 2026-12-14. The method is written out above so you can check our math.