Anthropic Claude Opus 5 Beats ChatGPT 5.6 Sol in ARC AGI 3

Anthropic’s Claude Opus 5 is being highlighted for outperforming ChatGPT 5.6 Sol and Fable 5 on reasoning and agentic benchmarks, including ARC AGI 3 and Frontier Bench.

The model is also said to run more cheaply on some workloads, such as a 3D city simulation.

If these results hold, Claude Opus 5 could become a stronger option for reasoning-heavy coding, simulations, and long-running AI workflows.