Grok 4.5 Scores 85.7% on ARC-AGI-1
Reports low per-task costs and results on harder variants.
Grok 4.5 from xAI posted verified scores on the ARC-AGI benchmark: 85.7% on ARC-AGI-1 at $0.33 per task, 52.6% on ARC-AGI-2 at $0.78 per task, and 0.3% on ARC-AGI-3. The results were shared via retweet by pseudonymous AI commentator @scaling01, who highlighted the model's scaling and capabilities. ARC Prize Foundation president Greg Kamradt replied positively to related claims about Grok 4.6 and stated readiness to test when available. These figures represent current confirmed performance on the public ARC benchmarks.
Combined views
653
2 posts, first seen 6h ago
Grok 4.5 Scores 85.7% on ARC-AGI-1
Reports low per-task costs and results on harder variants.
Grok 4.5 from xAI posted verified scores on the ARC-AGI benchmark: 85.7% on ARC-AGI-1 at $0.33 per task, 52.6% on ARC-AGI-2 at $0.78 per task, and 0.3% on ARC-AGI-3. The results were shared via retweet by pseudonymous AI commentator @scaling01, who highlighted the model's scaling and capabilities. ARC Prize Foundation president Greg Kamradt replied positively to related claims about Grok 4.6 and stated readiness to test when available. These figures represent current confirmed performance on the public ARC benchmarks.
Reactions from ranked influencers
2 posts
Grok 4.5 from @SpaceXAI on ARC-AGI (Verified) - ARC-AGI-1: 85.7%, $0.33/task - ARC-AGI-2: 52.6%, $0.78/task - ARC-AGI-3: 0.3%, $6.9K During testing we found that raising the reasoning effort from medium to high did not result in increased performance or cost
@elonmusk @arcprize @SpaceXAI I have no doubt. We're looking forward to it. Ready to test when you are @veggie_eric