What will my workload cost per month?

For each model the calculator bills one request the way our gateway does, then multiplies by the number of requests:

per request = input tokens  × input rate  / 1,000,000
            + output tokens × output rate / 1,000,000
              rounded to the nearest $0.000002, at least $0.000002
per month   = per request × requests per day × days

With the defaults (8,000 input and 600 output tokens) on GPT-5.6 Sol, one request costs 8,000 × $5.5 / 1M + 600 × $33 / 1M = $0.0638, and 6,000 of them cost $382.80. $0.000002 is the smallest amount the gateway bills (1 credit = $1 = 500,000 units), which is why a request is rounded before it is multiplied.

What it leaves out

• Prompt caching. Repeated context can be billed at a lower cached-input rate where the model supports it, so agent workloads often cost less than shown. Writing a prompt into the cache costs more than plain input when the provider reports it: 1.25× the input rate for those tokens, 2× for a 1-hour cache.
• Reasoning tokens. Models that think before answering bill those tokens as output. If you use a reasoning effort setting, raise the output figure.
• Agent loops. One task in a coding agent is many requests, and each re-sends the growing conversation. Count requests, not tasks.

Billing on Relaykit is prepaid: credits come in packs from $10, and each request is deducted from your balance using the token counts the provider reports. See how billing works, or estimate a single prompt with the token estimator.