Concept Country has developed its first open-source model, Qwen Concept 27B.

With 54% less thinking tokens and 3% better performance on GPQA Diamond compared to its base Qwen 3.8 model, we have achieved 30% more useful tokens per joule without additional post-training.

ConceptNov 623

Testing many models, many engines, many quants.

These are two exceptional outputs from everyone's favorite local model, Qwen 3.8 Flash. In all models, certain weights matter more than others. Some engines are fast because they round off the important bits. I think that's why

Mooncast ProductionsNov 63

Qwen 3.8 passed on the Hyperframes Slide to Claude.

Qwen doesn't have the power for... Draft quality.

H.G.O.D.Nov 5

I was shocked too. I was expecting three complex pages of jargon.

So I tried it with the model I had running at the time - Qwen-3.8-27B @ Q8: "build a floating pagoda island in three js"

Andrew WilsonNov 52