Thanks, I've totally missed that

It's basically the same as with the Qwen2.5 and 3 series but this time with 1M context and 200k native, yay :)