Opening soon
Private benchmarks are not open yet
We are still building the measurement side out. Leave an email and you will hear when it opens, and what we find in the meantime.
What you will actually get
New models, as they land
The category is days old and repositories are appearing faster than anyone is checking them. We track every one we can verify and say plainly which have been measured.
What the numbers survive
When a published figure does not reproduce, we say so and show the working. We have already hit one that does not.
When private benchmarks open
Your own historical decisions, frozen into a test set, run against the candidates. Not open yet. You will hear first.