Skip to content
● Deploy anywhereGet API key

Free · 14 pages · PDF · evidence checked 21 November 2026

Flat-Rate AI Budget Book 2027

A working workbook for teams that need to put a number on AI API spend before the invoice arrives — without building a per-token spreadsheet from scratch.

Get the PDF ↓

Short answer: the Budget Book is a 14-page PDF that turns your monthly request and token volume into a plan choice and a monthly cost, using the limits Plugsky actually enforces. It includes a worked example, a blank worksheet, three filled scenarios, a hidden-cost checklist and a one-page procurement summary. It is free; you confirm an email and download it.

What is inside the 14-page Budget Book?

  1. Cover — what the book is, and the date every figure was checked.
  2. How to use this book in 10 minutes — the ten-minute sequence and what the book is not.
  3. The forecast problem: why AI bills move — five drivers, the formula, and why the average is not the peak.
  4. Two billing models, one decision — flat rate and per token, side by side, with the three questions that decide it.
  5. Worked example: a support assistant at 20,000 requests a month — step by step to a verdict.
  6. The effective rate: what you pay per million tokens — the rate at different volumes, and the trap in a low one.
  7. The budget worksheet — formulas, inputs, results, and what to watch.
  8. Three worked worksheets — support chat, RAG assistant, agent workflow, with sensitivity checks.
  9. Hidden costs I: retries and context — the two that break most estimates.
  10. Hidden costs II: embeddings, evaluations, storage, media, seats — the costs outside the token bill.
  11. The decision tree: choose the smallest plan that fits — five questions to a plan.
  12. The one-page procurement checklist — the page you attach to an internal request.
  13. What to watch: fair use, concurrency and the peak day — where a flat plan stops being cheaper.
  14. Sources and methodology — every figure dated.

The plan facts the workbook uses

Published Plugsky plan limits, checked 21 November 2026. Fair-use concurrency and RPM apply on every plan.
PlanPriceRequests / dayTokens / dayConcurrencyRPM
Free$0500 (100 for new accounts)250,000 (50,000 new)2 (1 new)30
Hobby$8/mo1,0002,000,0004120
Starter$20/mo2,0005,000,0006240
Builder$60/moUnlimited — no token capUnlimited — no token cap10600
Scale$120/moUnlimited — no capUnlimited — no cap161,200

Annual billing is 20% below twelve monthly payments. “Unlimited” on Builder and Scale means no monthly token cap and no per-token overage; concurrency and RPM still apply.

How does a flat rate change forecasting?

On a per-token bill the forecast is a multiplication problem with three uncertain inputs: tokens per request, requests per month, and the price of the model you happened to route to. Every one of those drifts. Prompts grow because someone adds a worked example; retries double a request without anyone noticing; a fallback model can be several times the price of the primary. The arithmetic is simple and the answer is wrong by the second week.

A flat plan removes two of the three inputs. The price is fixed for the month, and the only question left is whether your volume fits inside the plan's daily caps. That turns a forecast into a capacity check: requests per day, tokens per day, and peak concurrency. If the volume fits, the bill is the plan price. If it does not, the workbook tells you which cap breaks first — and that is the number to take to a plan upgrade conversation rather than to a finance argument.

It is not free of risk. A flat plan is cheaper only while your usage sits below the cap you bought. Above it, the economics invert and a per-token bill would have been lower. That is why the workbook includes a watch-outs page instead of ending on a recommendation.

Worked example: one team, one month

Illustrative volumes, not a customer measurement. Arithmetic uses the published limits in the table above.
InputValueWhat it decides
Average requests per day1,400Hobby's 1,000/day cap is exceeded; Starter's 2,000 fits
Average tokens per request2,8003,920,000 tokens/day — inside Starter's 5,000,000 cap
Peak concurrent calls5Starter allows 6
Resulting planStarter$20/month, 20% less on annual billing

The same volume on Hobby would fail on the request cap before it ever reached the token cap. That ordering matters: teams usually watch tokens and are surprised by the request limit. The workbook checks both, in the order the API enforces them.

Effective rate per million tokens

Comparing plans is easier when the price is expressed per million tokens at the daily cap. These are derived from the published limits and a 30-day month, and they are the figures printed in the workbook's rate table.

Derived from plan price and daily token caps; assumes the cap is fully used every day.
PlanMonthly priceIncluded tokens / monthEffective rate / 1M tokens
Hobby$860,000,000$0.133
Starter$20150,000,000$0.133
Builder$60No monthly capNot comparable
Scale$120No monthly capNot comparable

A lower effective rate is not automatically better: it assumes you can use the cap. A team at 20% utilisation pays the same price for tokens it never sends.

Who should not use this workbook

Who is the Budget Book for?

How do you use the workbook?

  1. Collect three numbers — average requests per day, average tokens per request, and peak concurrent calls.
  2. Fill one row of the worksheet. It returns the smallest plan whose daily caps cover the month.
  3. Read the hidden-cost pages before signing off, because retries and context resend are what turn a comfortable estimate into an uncomfortable invoice.

What does the Budget Book not cover?

It is a planning workbook, not a billing statement. It does not predict your retry rate, your prompt-growth over time, or the cost of evaluation runs you have not designed yet. It does not include competitor prices, because most are quote-only and change without notice. And it cannot tell you whether your workload is a good fit for a flat plan at all — the watch-outs page is where that argument lives.

Get the PDF

One email to confirm, then the download link. No spam. Unsubscribe anytime.

Questions

Is the Budget Book free?

Yes. Enter a work email, confirm the link we send, and the 14-page PDF downloads immediately. No card is required.

Which plan limits does it use?

The limits Plugsky actually enforces today: Hobby at 1,000 requests and 2,000,000 tokens per day, Starter at 2,000 requests and 5,000,000 tokens per day, and Builder and Scale with no monthly token cap under fair use.

Does it include competitor pricing?

No. The workbook uses Plugsky's published plans only. Competitor prices change and are often quote-only, so including them without a dated source would make the arithmetic unreliable.

Is the forecast a guarantee?

No. The worksheet multiplies your own monthly volume by the plan price. It cannot predict retries, context resend or evaluation runs, which is why the hidden-cost checklist exists.

Can I use it in a procurement pack?

Yes. Page 12 is a one-page procurement checklist you can attach to an internal request. Every figure is dated 21 November 2026 so a reviewer knows which version they are reading.

What happens after I download it?

You join the monthly AI Cost Index list, which includes occasional product updates from Plugsky. Unsubscribe from any email, any time.

Related

Sources