Tier 1 · instant download
LLM cost model spreadsheet
Model the spend before you commit to an architecture that sets it.
What you get
- Model routing
- Cache hit rates
- Quantization effects
Three ways to tell whether this applies to you.
If none of these describes your situation, this is the wrong product, and the catalog has nine others.
- You are about to commit to an architecture that will set your model spend
- You need to show someone the cost before the invoice shows it to you
- You want to model routing, caching and quantization rather than guess at them
The file, before you buy it
Read from the artifact itself rather than described, so what is printed here and what arrives cannot drift apart.
cost-model-guide.html
7 sections, in the same download.
- What is in the spreadsheet
- Lever 1 · Model routing
- Lever 2 · Cache hit rate
- Lever 3 · Quantization, and self-hosting
- The order to pull the levers
- What this model deliberately leaves out
- Sources
Three levers, in order
The cheapest model is the lever most teams reach for first, and it is rarely the largest saving.
The guide names what the model deliberately leaves out, which is the part that makes the rest usable: a cost model claiming to cover everything is one nobody can check.
- 01Model routingRouting sends each request to the cheapest tier that can answer it, escalating when it cannot.
- 02Cache hit ratePrompt caching discounts input tokens that repeat across requests.
- 03Quantization, and self-hostingQuantization reduces the numeric precision of a model's weights, cutting memory and raising throughput per GPU.
SourceSection headings read from deliverables/T1-ENG-03/cost-model-guide.html at build time.
Delivery and license
Stated here rather than in the terms, because these are the questions asked before a purchase and not after it.
- Delivery
- Immediate download.
- License
- Single organization, internal use, perpetual. Adapt it for your own systems freely. Do not resell it or publish it as your own.
HTML and CSV · Spreadsheet with live formulas, plus a guide to the three levers · Version 1.0 · 21 September 2026
What usually goes with this
Each of these does something this one does not, and the line under it says what. They are separate items, not an upsell path.
RAG architecture decision guide
The decisions behind the numbers this model moves: chunking, embedding choice, index type, reranking.
$290
Hallucination evaluation starter kit
Routing part of your traffic to a cheaper model is only safe if you are measuring what it does to groundedness.
$490
Before you ask
These are the same strings this page puts in its structured data, so a person and an answer engine read identical text.
Can I buy this today?
Yes. It is $149, one payment, in USD, and the file downloads as soon as the payment clears.
What if I want the work done rather than the file?
The fixed-scope engagements on the services page cover the same ground with a person doing the work, at a published price and a fixed duration. Those are available now: a vendor AI assessment at $2,500 over three days is the smallest, an EU AI Act readiness assessment is $4,900 over one week, and an ISO/IEC 42001 gap assessment is $8,900 over two to three weeks.
What exactly is LLM cost model spreadsheet?
Model the spend before you commit to an architecture that sets it. It is a Tier 1 instant product from MetaMinds, listed at $149 in USD as a single payment. Tier 1 is the entry rung of a three-tier catalog: instant products from $99 to $890, fixed-scope engagements from $2,500 to $8,900, and scoped builds from $18,000.
What happens after I pay?
The download link and your invoice are emailed to you immediately, and the link stays good for 30 days and up to 20 downloads. Because a file you have already downloaded cannot be returned, the checkout asks you to confirm before paying that downloading ends your 14-day right to withdraw. That is Article 16(m) of the Consumer Rights Directive, and it is why the box is there and why it starts unticked.
It downloads the moment you pay.
No call, no scoping, no waiting. If you would rather a person did this work, the engagements on the services page cover the same ground at a fixed price and a fixed duration.