You can run it on high end laptop at like 0.1 token per second.

But I think smaller size models like qwen 3.8 are way more important.