Qwen 3.8 27B now runs locally inside RunAnywhere!
Your app can use the same local AI setup across different hardware. On Mac it runs Swift + mlx. On Windows, Electron + llama.cpp. Different architectures, different inference engines, one SDK call. Your application code doesn’t
A 27B model you can run on a single GPU that matches GPT-5.6 Luna?
Artificial Analysis says yes. Our benchmark says… let's talk. Qwen 3.8 27B landed with a real claim behind it: Artificial Analysis scores it 52 on its
Qwen 3.8 Max beat Fable 5 at building 3D physics scenes for 7x cheaper!
Qwen 3.8 Max is now being served in Nexum Router by with 10x more useage boost including all models.
Pricing starts just at $5.5 per week only.
You have reached the end of the archive
All of qwen38