GPT-6 Astra Hits Record 62.7% on ARC-AGI-3

GPT-6 Astra scored 62.7% on the standard ARC-AGI-3 harness—about double Opus 5—and reached 99.9% with a provider adapter harness, shattering records on a benchmark where models previously struggled to exceed 10%.

2026-09-04 ~ 2026-09-04 · 3 related posts