gpt-6VIDEO

GPT-6 Sol falls short of Opus 5.5 on a Three.js frontend test

Given the same New York scene prompt, Opus 5.5 built a full 3D scene with a movable camera, while GPT-6 Sol produced a single static frame with no render loop in its code.

aipulsedaily

@aipulseda1lyArchived 2026.09.23Originally posted 2026.09.23

aipulsedaily @aipulseda1lyOriginally posted 2026.09.23

Open original on X

OpenAI has no chance against Opus 5.5 in frontend and Three.js. Opus 5.5 hit everything in my NYC prompt and went above and beyond. Street level with taxis, rooftops, the Brooklyn Bridge, Central Park. GPT-6 Sol gave me one static frame. You cant move the camera. Its code doesnt even have a render loop, it draws once and stops. GPT-6 Sol on top, Opus 5.5 below. Full run, nothing cut.

x.com/aipulseda1ly/status/2102465200666370514Archived 2026.09.23 10:17