Qwen dropped a qoduck model, which is currently free.
After doing some comparisons, 30 different texts all matched the Qwen3.8 word list, and the visual counts also matched. Belongs to the Qwen series model and supports 1M context Either qwen-3.8 or qwen-4
I threw 20 HARD tests at Qwen 3.8 27B on my own PC...
And it nailed 19 of them 🤯 19 passed, 1 partly. The only slip was a polite email rewrite that added dollar signs nobody asked for 👀 Full test below 👇
Everyone saw the "goose riding a bicycle" prompt going around.
I decided to see what happens when you dial the difficulty up to 10: A peacock on a bike. Prompt: "run 2 subagents to create 2 unique peacocks riding a bicycle animated svg" Instead of hitting a cloud API, I
Qoder’s new anonymous model Qoduck VS Qwen 3.8 max 3D voxel floating pagoda comparison
This Qoduck is probably not a large parameter model. The quality of the generated content is a bit strange, and it even feels like it’s not well trained. Not as good as Qwen 3.8 MAX, let alone Qwen 4
You have reached the end of the archive
All of qwen38