Which AI is best at what?
Rankings from millions of blind votes on Arena, where people compare two anonymous answers and pick the better one.
The Arena Orbit
Each spoke is a category. The closer a dot sits to the outer ring, the higher that company's best model ranks there. Each company's dots form its shape.
Loading…
Reign of #1
Who held the top spot in this category, update by update.
Head to head
Pick any two models and compare their rank in every category.
How to read this
Scores are Elo-style ratings. A 20-point gap means the higher model wins a bit over half of head-to-head votes. Gaps under ~10 points are often within the margin of error.
Each model appears once. Labs often enter several settings of one model (like High and Max effort). We show the best-ranked setting so the list isn't crowded with near-duplicates.
Source and freshness. Data from Arena's public dataset, refreshed automatically. Each category shows the date Arena last published it.