the benchmark

How it compares.

Every number in the first column is zero. We checked twice.

the ai pebble, measured against a large language model
Measure the ai pebble a large language model
Energy per query0 Wnon-zero, rarely published
Water per query0 mLnon-zero, rarely published
Hallucination rate0%greater than zero
Cost after purchase$0, foreverusually recurring
Data collected about younonesee their privacy policy
Training datanonenot always disclosed
Context window0 tokenslarge, and growing
Model updatesnone, everfrequent
Times it has been confidently wrong0at least once
Works with no internetyesno
Works during their outageyesno
Account requirednoyes

The first column is verifiable. Pick it up and check. The second column is deliberately vague, partly because we are not going to name anyone, and partly because for several of these rows the honest answer is that they do not say.

We are aware that winning this comparison requires doing nothing at all. That is the product.

It wins by not competing.

A rock cannot hallucinate, cannot be trained on anyone's work without asking, cannot leak what you told it, and cannot be discontinued when the funding runs out. None of that is engineering. It is just what happens when a thing does nothing.

buy · $20