Claude was unsubscribed as ChatGPT/Codex 5.6, Sol and Kimi 3 all struggled

sull · x · 2026-07-28

The author says they unsubscribed from Claude and found that ChatGPT/Codex 5.6, Sol, and Kimi 3 all struggled on hard tasks that day.

Their conclusion is blunt: we are not there yet. It is a short, subjective capability check rather than a full benchmark, but it clearly points to model performance.

Original post →

More from Models

Models channel →