Opus 5.5 vs. GPT-6 Astra vs. Sonnet 5.5 vs. GPT-6.1 Sol. I gave all four the same one-line prompt: recreate Sonic in Three.js. No reference image. No follow-ups. Just one prompt. Who wins?

GPT cooked for blender. Nothing beat them tbh

The first kind of test that OpenAI finally wins

sol looking that clean for pocket change is nuts 🤣

Wait, did Opus really created that with that much money spent? 🤯

6.1 sol 💪🌝

guaranteed at least one of them generated a blue sphere with limbs

Look at the prices

Look at the cost line under each one: GPT-6.1 Sol took 11 min and $0.54, Sonnet 5.5 took 44 min and $8.14. Astra's face looks closest to Sonic to me, but Sol is cheapest by a mile.

GPT-6.1 Sol winning on both quality and cost is a pretty wild result. The fact that all four got the exact same one-line prompt makes this comparison way more interesting.

Oh look, when the goal is actually clear (all the models can search up the result) GPT does better. Opus vibes harder, GPT actually builds what you ask for.. no more, no less.. aka it does work not vibes.

You are cheating. Because what you are not saying is that Claude rejects always recreate trademarks. But he asks if you want something similar but clearly different. You said YES to something different. Happened to me too...

Astra and 6.1 sol ☺️

they all look pretty bad, to be honest. Expected better from Astra. I think Mario came out better overall because he is a simpler design than Sonic.