OpenAI's Luna Aces ARC-AGI-1 at 90.7% with Massive 80% Cost Reduction

burny_tech · x · 2026-08-07

ARC-AGI re-tested OpenAI's GPT-5.6 Luna following an 80% price reduction. The model scored 90.7% on ARC-AGI-1 at just $0.07/task, and 59.6% on the harder ARC-AGI-2 at $0.18/task.

The new results match the model's original performance at a fraction of the cost. This highlights a massive leap in inference economics compared to the $20,000 previously spent for o1-Preview to hit 80%.

Related event: OpenAI Luna Hits 90% on ARC-AGI After Price Cut(2 posts)→

Original post →

More from Models

Models channel →