DeepSeek V4.1 Flash vs MiniMax M3.1 on procedural turbofan 3D assembly

Same prompt, two procedural turbofans: DeepSeek V4.1 Flash 1017 vs MiniMax M3.1 Flash compared on explode-hold-reassemble spatial math.

Fazley | AI Lab

@itsfazleyArchived 10 days agoOriginally posted 2026.10.21

Fazley | AI Lab @itsfazleyOriginally posted 2026.10.21

Open original on X

One prompt. Two procedural turbofans. Zero textures. 🔺 TOP: DeepSeek V4.1 Flash 1017 (247K tokens) 🔻 BOTTOM: MiniMax M3.1 Flash (227K tokens) Explode → hold → reassemble. Which spatial math handled the assembly better? Top or Bottom? 👇

x.com/itsfazley/status/2112914304060252404Archived 2026.10.22 12:00

Open original on X