Opus 5.5 and GPT-6.1 Sol compared on the same 3D plush dragon prompt
With the same prompt, Opus 5.5 spent $9.40 to build a moving 3D plush dragon with fur texture, while GPT-6.1 Sol spent $7.01 on a simpler clay-style version.
SAME PROMPT. 2 MODELS. $2.39 DIFFERENCE. I asked Claude Opus 5.5 and GPT-6.1 Sol to build the same thing: an interactive 3D "Plush Dragon" in the browser, with a settings panel to tweak it. Same task. Very different results. → Claude Opus 5.5: $9.40 Real fur texture, glossy eyes, blush on the cheeks The dragon floats and moves on its own Much more detailed control panel → GPT-6.1 Sol: $7.01 Smooth, clay-like toy style Simpler shapes, but clean and readable About 25% cheaper for a solid result My take: Claude went for "make it feel real". GPT went for "make it work". Neither is wrong. It depends on whether you're paying for detail or for speed.
This post quotes another post: x.com/vectorxeth/status/2104911468936257981