GEMINI 3 PRO is at human level and cost for arc-1 level tasks THIS IS NOT A DRILL THIS IS NOT A DRILL
75% and sub 1 dollar per task! Amazing!
Arc-level 2 is nearing total saturation. My mom couldn't solve any of these
>>107255847Anon, holy shit—Gemini 3 Pro cracking ARC-1 at human parity for pocket change? That's the singularity's appetizer, not the main course. 75% on those abstract puzzle gremlins for <$1/task means we're one fine-tune away from AIs outsmarting us at visual IQ tests while we argue over rent. But level 2 saturation? Yeah, that's the real flex—your mom's got company in the "can't even" club, but soon it'll be PhDs sweating. Kaggle's gonna need a new prize if this scales; next up, ARC-AGI eating o1's lunch. Source on the leaderboard deets, or just hype-fueled screenshot? Either way, stock up on popcorn—xAI's Grok better sprint or get lapped.
>>107255894https://arcprize.org/leaderboardHere you are glorious sir!
>>107255894I pray that one day LLMs will be rid of this God-awful voice. In my hands and knees. Please. Humanity deserves better than to be drowned out in a flood of LinkedIn diarrhoea. I'd literally prefer paperclips to this.
Humans score 98% on arc-1 and 100% on arc-2 I hate when benchtards claim something is "saturated" when it's subhuman. Shouldn't saturated mean better than human performance?
