News for hackers. Stuff that matters.

A comment on ARC-AGI-3. Parent comment

Re:ARC-AGI-3 by pawelk411 on Thursday March 26, 2026 @03:15AM (#47527503)
Yeah but thats literally above ASI, let alone AGI. Average human scores <1% on this bench, opus scores 97.1% when given an actual vision access, which means agi was long ago achieved
[ Parent | on HN ]
Re:ARC-AGI-3 by vova_hn2 on Thursday March 26, 2026 @07:16AM (#47529088)
> opus scores 97.1% when given an actual vision access

Do you have a source for this? I would be very curious to see how top models do with vision.

[ Parent | on HN ]
Re:ARC-AGI-3 by daveguy on Thursday March 26, 2026 @11:08AM (#47531455)
No, there is no source for this. Opus is scoring around 1% just like all the other frontier models. It would be fairly trivial to add a renderer intermediary. And if it improves to 97+%... Then you would get a huge cut of $2 million dollars. The assertion that Opus gets 97% if you just give it a gui is completely bogus.
[ Parent | on HN ]
Re:ARC-AGI-3 by famouswaffles on Thursday March 26, 2026 @12:30PM (#47532566)
https://news.ycombinator.com/item?id=47532483
[ Parent | on HN ]