AI Model Cost Audit

How the audit works

1. 1

   Book a call

   Pick a time on the booking page. On the call we agree which AI features and models the audit covers.

2. 2

   Share your usage

   Send your usage export or invoices and a few sample prompts. We need no access to your systems.

3. 3

   We review it

   We map each call to its task, match the task to the independent benchmarks that measure it and price the models that score as well for less.

4. 4

   Report and walkthrough

   You get the shortlist for each task, the savings estimate with every assumption written down and the quick wins your team can apply on its own, then a 30-minute call with the founder to walk through it.

Where the savings come from

The report looks for two kinds of saving: a cheaper model for the tasks that do not need a frontier one, and changes to how you call the models you keep.

A cheaper model for each task

Classification, extraction and short summaries often score as well on a cheaper model. The shortlist names the models that match each task on the benchmarks that measure it.

Prompt caching

Instructions and reference documents sent again and again can be billed at the cached rate, a fraction of the regular input price.

Batching

Work that can wait, such as overnight document runs, can go through a batch API, which bills at a lower rate.

Shorter prompts and answers

Output usually costs several times as much as input, so trimming prompts and capping answers lowers the bill directly.

Routing rules

Rules that send each request to the cheapest model that handles it well, and pass only the hard ones to a frontier model.

Where the shortlist comes from

The shortlist uses the data behind our AI model benchmarks: independent benchmark scores and current list prices for more than 140 models, each with its source and the day it was checked.

Labs’ own launch tables are shown apart and never rank a model. Public scores are a starting point, so try the shortlisted models on your own prompts before you switch.

Questions about the audit

What does the AI Model Cost Audit include?

A review of your current AI spend by model and task, a shortlist of cheaper models for each task that score as well on the benchmarks that match it, a savings estimate with every assumption written down, quick wins your team can apply on its own and a 30-minute call with the founder to walk through the report.

What do you need from us?

Your usage export or invoices and a few sample prompts. We need no access to your systems.

How do you know a cheaper model is good enough?

We match each task to the independent benchmarks that measure it and shortlist the models that score as well there for less. Public scores are a starting point, so try the shortlist on your own prompts before you switch.

Which models do you compare?

The models in our AI model benchmarks: more than 140, from Google, Anthropic, OpenAI and other providers, with independent benchmark scores and current list prices where they exist.

Is the audit part of a larger project?

No. It is a one-time, fixed-price review on its own. If you want help applying the changes, our AI agent development service builds the routing, cost controls and evaluation around the models you choose.

See what your AI bill could be

Book a call, send your usage export or invoices and get the report, with a walkthrough from the founder.