Press Space to continue
Claude models vastly improve on ARC-AGI benchmark tasks
Press Space to continue
GPT 6 ASTA obliterated the ARC-AGI-3. when it was launched models struggled to get to 10%. now Astra scores: > 62.7% with the standard harness > 99.9% with provider adapter harness
open on x · 1 viewViews and clicks count once per visitor per day. Clicks are opens of this post on X.