And on a real phone? No dgx no GPU only CPU?
Imagine running Qwen 3.8 Flash Next on a 12GB Android phone
And phone. Imagine running Qwen 3.8 Flash Next on a 12GB Android phone - 2 tok/s
Imagine running Qwen 3.8 Flash Next on a 12GB Android phone
CARNICE-V3-27B is a 27B local AI model built to act, not just chat.
But the most interesting part isn't the model size. It's how little training went into creating it. What makes Carnice different: → 27.8B parameters built on Qwen 3.8-27B → Specifically trained for agentic
You have reached the end of the archive
All of qwen38