Estimate Product Plan output units and cost

Use traffic, answer length, and cached input to estimate monthly output capacity and compare plan cost with token charges.

By Standard Thinking · Updated

Planning figures checked . Product Plan prices and allocation terms are confirmed before paid use. One unit is one model's base allocation of aggregate output tokens per second — GLM-5.3 at 135, Kimi K3 at 80, DeepSeek V4.1 Flash at 550 — at $1,200 per unit, per model, per month, with auto scaling at $5 per extra unit per 2-hour block.

Size a plan for your traffic

USD · 30-day month · No API requests are made

Your traffic

An assumed share of input tokens, not a predicted cache hit rate. It moves the pay-per-token side and the break-even figure, never the plan price.

Estimated monthly cost

1,125,000 answers a month · 1.8B input · 450M output tokens

Product Plan · GLM-5.3

Base capacity$2,400
Auto scaling$300
Total$2,700

2 units × $1,200 a month · 1 more unit × 60 blocks × $5

Standard API · pay per token

Input$2,520
Output$1,980
Total$4,500

1.8B input at $1.40 and 450M output at $4.40 per million tokens · 0% cached input

Normal-hour rate

What an ordinary hour asks of the plan

138.9 tokens/s

Peak-hour rate

What the busiest hour asks

347.2 tokens/s

Estimated use of base output capacity

Output assigned to the base, divided by its monthly output capacity

60%

Break-even use

Where one unit's pay-per-token cost reaches its monthly price

34%

Answers/hour at listed base output rate

Arithmetic capacity at this answer length; input processing and latency are not modeled

2,430

Pay per token would cost $1,800 more a month for this traffic.

Planning figures checked . Product Plan figures are indicative; confirm current prices and allocation terms before paid use. Every result is calculated from your assumptions, not measured traffic.

Indicative plan terms and API rates

One unit's output rate per model, and the Standard API rates used for the pay-per-token side.
ModelOne unitInputCache readOutputSource
GLM-5.3Unit source135$1.40$0.26$4.40Rate source
Kimi K3Unit source80$2.55$0.255$12.75Rate source
DeepSeek V4.1 FlashUnit source550$0.24$0.0048$0.96Rate source

Unit rates in aggregate output tokens per second · token rates in USD per 1 million tokens · $1,200 per unit, per model, per month · $5 per extra unit per 2-hour block

Choose inputs that describe your traffic

This calculator estimates Product Plan output units and monthly cost, then compares the same requests with Standard API token pricing. It sizes output demand from average hourly traffic; it does not validate the complete workload or guarantee response times.

Select a model, then enter:

  • Answers in a normal hour: the baseline rate used to size monthly base units.
  • Answers in a peak hour: the total rate during a peak, including the baseline traffic.
  • Peak hours per day and peak days per month: how often that higher rate replaces the normal rate. The estimate treats those hours as one daily peak period.
  • Average input and output tokens per answer: count the model calls behind the feature, including agent turns and retries where applicable.
  • Cached input share: the percentage of input tokens expected to be billed at the selected cache-read rate. It is not the percentage of requests that use caching.

The month length, unit rates, prices, and block length come from Indicative plan terms and API rates. Reset restores the illustrative traffic assumptions, rather than observed customer usage. If you have no measurements yet, the token estimator can help build an initial call-volume assumption.

How to read the result

Base units round the normal hour’s output demand up to whole units, with a minimum of one. Even zero traffic therefore retains a minimum plan cost. Auto scaling covers the difference between peak units and base units, multiplying it by billed blocks and the block price.

Peak hours are rounded up to whole blocks each day. A one-hour peak and a two-hour peak can incur the same block charge. Separate bursts may need a different count depending on their timing and the agreed billing rules; this tool does not reconstruct those events from a traffic trace.

Pay per token uses the same total answers and token lengths. It applies ordinary and cached input rates separately, plus the output rate. Changing the cached share changes this estimated API bill; it does not change the output-based unit count.

Use of the base measures output assigned to the base over the month, capped at its rate in each traffic period. Output above that rate belongs to extra capacity and is excluded from base use. Break-even use compares one unit’s monthly price with its theoretical token bill at continuous full output, using your input-to-output ratio and cached share. Compare the complete monthly bills to make a decision: the break-even percentage alone does not include auto scaling charges.

The calculation

normal rate = normal answers per hour × output tokens ÷ 3,600
peak rate   = peak answers per hour × output tokens ÷ 3,600
base units  = max(1, ceil(normal rate ÷ unit rate))
peak units  = max(1, ceil(peak rate ÷ unit rate))
extra units = max(0, peak units − base units)
blocks      = ceil(peak hours per day ÷ block hours) × peak days
plan cost   = base units × monthly unit price
            + extra units × blocks × block price

peak hours    = min(month hours, peak hours per day × peak days)
normal hours  = month hours − peak hours
monthly calls = normal hourly calls × normal hours
              + peak hourly calls × peak hours
API cost      = (uncached input × input rate + cached input × cache rate
              + output × output rate) ÷ 1,000,000

Token totals are monthly calls multiplied by average tokens per call. Calculations retain unrounded values; displayed figures are rounded. If average output is zero, an answers-per-hour conversion is undefined; an input-only workload still requires a separate capacity assessment.

What this calculator does not decide

The estimate does not size input processing, concurrency, short bursts, or latency. A workload can fit the calculated hourly output rate and still require different service conditions. See output tokens per second explained for an example of the difference between average demand and a burst.

Free extra capacity from the shared pool is excluded because it is not reserved. The tool prices extra units as authorized auto scaling, which remains subject to availability. It compares one model at a time and excludes the three-model bundle. It also excludes taxes, credits, negotiated discounts, tool execution, storage, networking, and other application costs.

Listed prices are indicative and terms are confirmed before paid use. Request access with your traffic measurements, input lengths, and latency requirements to assess the estimate against an actual allocation.

Sources and verification

Sources used for this resource. Verification dates describe our source checks, not provider effective dates.

Standard Thinking Product Plan

Listed Product Plan allocations and indicative prices are planning assumptions, subject to order terms. Your order defines the model, allocation, measurement conditions, pricing, and any service commitments. Product Plan inference is covered by ZDR when activated.

Checked

Standard Thinking pricing

Standard API rates in USD per 1M tokens used for the pay-per-token side of this calculator, re-read on October 1, 2026 and unchanged: GLM-5.3 $1.40 input, $4.40 output, $0.26 cache read; Kimi K3 $2.55, $12.75, $0.255; DeepSeek V4.1 Flash $0.24, $0.96, $0.0048.

Checked

View all resources