Claude Opus 5 outbuilds GPT-5.6 Sol on a shared Three.js extreme-ocean prompt, at nearly 10x spend

In a same-prompt coding trial, Claude Opus 5 produced a substantially more ambitious extreme-weather ocean scene in Three.js than OpenAI’s GPT-5.6 Sol, while spending almost 10 times as much and continuing to expand the work without any loop instruction. The comparison, run by a developer who first asked Sol to build a violent, storm-driven ocean entirely in Three.js and then gave Opus 5 the identical challenge, has become a practical illustration of how the two flagships differ when left to long-horizon visual and simulation work.