Four models. Three tasks.
ZERO FIXES. everything looked the same at the start. but after one prompt, the difference became obvious. GPT-5.6, Claude Opus 5, Qwen 3.8 Max and Grok 4.6 had to build a premium product page, a playable browser game, and an animated rainy city scene
Qwen 3.8 27B is pretty good with canvas, drawing with falling sand, 2:38 was a masterpiece, lol
Qwen 3.8 Max just beat Anthropic's Fable 5 at 3D physics generation for 7x less cost.
There's one catch. Both models got the same task: build three self-running 3D scenes with real physics, a marble machine, an assembly line, and a sawmill simulation. Qwen's textures,
Qwen 3.8 27B ROCmi4 running in Hermes being served by ROCmFPX build.
Real world example of chat. TTFT can be improved but for a dense model and it being a new release I'm happy. The responses 🚀 Hardware: FrameworkPuter AMD Strix Halo Software: ROCmFPX build Agent: Hermes
You have reached the end of the archive
All of qwen38