AI Intelligence
Benchmarks are facts.
DNA is judgement.
Everyone publishes the scores. Almost nobody tells you what they mean for the work you actually do. We give you both: the raw numbers as the labs published them, and our own rating across the eight axes that decide which model you should reach for.
Frontier leaderboard
Ranked by Solthar DNA, the average of our eight axis ratings. Sort by any column to rank the way you care about.
DNA is Solthar's own assessment, not a benchmark. For published eval scores, see Test scores.
Model DNA
Eight axes, rated 0–5. Not a benchmark aggregate. Our read on what each model is actually good at, so you can match a model to a job instead of to a scoreboard.
What the eight axes mean
Every model, side by side
Benchmarks
Benchmarks measure benchmarks. A two-point gap is usually noise; a ten-point gap is usually real.
AI Arena
Two models, every axis, side by side. The winner of each row is highlighted. And the overall call at the bottom is derived from the data, not written by hand.
Model timeline
Every release we track, newest first. Assembled from each model's own history. Nothing here is typed twice.