DeepSeek V4.2 vs GPT-6 Luna: Same Minecraft Prompt — Which Result Looks Better?

I gave DeepSeek V4.2 and GPT-6 Luna the same prompt to build a Minecraft-style game in JavaScript, then compared the results side by side. Prompt: “Make an extremely realistic clone of Minecraft. Make it really good and exactly the same. Use everything you know, leave nothing out. Use JavaScript, and make it extremely realistic.” Which result looks more convincing to you? What details stood out?