Ask a specific question.
Start with a task worth understanding. Make it clear what a benchmark measures and what it leaves out.
An independent benchmark project
Better questions.
Clearer comparisons.
A place to explore what AI models can do, through personal benchmarks and the details behind each result.
Explore the benchmarksFirst benchmarks coming soon. Every result will come with the method behind it.
FIELD NOTES001 / THE BEGINNING
Every comparison starts
with a baseline.
The questions come first. The scores follow.
THE BENCHMARKS
An open notebook.
WORK IN PROGRESS
Read the approach The first questions are taking shape.
No results have been published yet. This is where you’ll find the benchmarks, the models tested, and what the results mean.
THE APPROACH
A score is a starting point.
Context makes it useful.
This project starts with curiosity about how different models handle specific tasks. The benchmark design comes next, with these principles in mind.
Show the working.
Give results their context: the test setup, model versions, and how the responses were evaluated.
Leave room for nuance.
Look beyond a single ranking. Explore the differences, limitations, and examples behind the numbers.