Claude Opus 5 aces 20x20-digit multiplication eval, 1200/1200 correct with max reasoning

maksym_andr · x · 2026-09-17

Developer maksymandr ran a minimal eval testing whether frontier LLMs can multiply large numbers without external tools. Claude Opus 5 scored 100% on multiplications up to 20x20 digits (1200/1200 correct), but only at maximum reasoning effort.

The author argues this simple eval is newly relevant for measuring no-CoT capability, especially for recurrent-depth architectures like GPT-6-Astra, which reportedly cannot run without CoT at all.

Related event: Claude Opus 5 Aces 20×20-Digit Multiplication, 1200/1200 Without Tools(7 posts)→

Original post →

More from Models

Models channel →