@kimmonismus: benchmark context for Carbon's 'number one'

@kimmonismus points out that Artificial Analysis' index, which Google cites for Carbon, now runs largely on private, held-out tests (he says about 60%), which makes it harder to game; he was also surprised Carbon topped it. He also flagged NVIDIA's MotionBricks for real-time character motion in games and robots (NVIDIA's figures; niche).

Why it mattersPrivate test sets make benchmark claims a little more trustworthy, but a one-point lead is still within the noise. Test on your own tasks.

Curated by Sara Simeone as part of our human editorial coverage. The original sources are linked below; curation is distinct from a Hype or Fact? assessment.

How we knowCompany figures are labelled as their own claims; anything we couldn't verify is labelled as a report or a claim.