A Qwen 3.8 27B on one Mac Studio beside DeepSeek on my four box Spark cluster.
Same harbor prompt. Qwen: 53.57 tokens/sec. DeepSeek: 107.70. IMO, both models are doing phenomenally with good fine tuning. Both demos had two correction rounds plus manual visual repairs. Speeds
For those who said that the stealth model QoDuck on qoder_ai_ide is not as good as Qwen 3.8 Max, it's simply not true at least with this test.
Same prompt and thinking level. Qwen 3.8 Max on left. QoDuck (Qwen 4) on right.
According to this blogger, his small Arctic theater was made with Qwen 4.
According to this blogger, his small Arctic theater was made with Qwen 4. Qoder just released the QoDuck model today. I tried the effect using the same prompt words. It seems that the effect is much different from Qwen 4. Even the 4 series is a small parameter model, right? However, this magnification is 0.5x, which is the same as the previous generation Qwen 3.8 Max, so this time Qwen 4 Max…
Floating Pagoda Showdown! QoDuck vs Qwen 3.8 Max, you can see the model texture at a glance!
✧ Comparison using the same 3D voxel floating pagoda question: Qwen 3.8 Max handed over a single Lingxiao pagoda with a blue night sky and a clean structure; QoDuck showed a group of floating island pagodas under an orange-red sky, with scattered particles and messy stacking of voxels. ✧…
You have reached the end of the archive
All of qwen38