Qwen 3.8 flash next built this landing page in 29 minutes and served it on my tailnet on its own.
I am running official fp8 on 2x dgx spark, full 256k context loaded, 45 tok/s with mtp on. i have run deepseek and glm on these boxes and it was good, this one just feels right.
Here's another good 3D object made by DeepSeek-V4.1-Flash.
This time a submarine/ underwater explorer. This is a 3D model view. No model file is being loaded here. This is procedurally modeled in code with Three.js. Zero GLB/OBJ assets. Zero downloaded models. No iteration,
5000 live physics particle simulation across static and dynamic spatial force vector fields.
Created by DeepSeek 4.1 Flash using We make authoring experiences like this feel natural for AI agents: think only in semantic authoring of forces, motion and
GTP6-Astra + Luna + DeepSeek V4.1-Flash
You have reached the end of the archive
All of deepseek