Jahaz/

Qwen-Coder-NX-73B

A Flagship 73B model, like what we remmeber in the past 3 years, lands in MoE.

This model base on Qwen-Coder-Next, which is my fav, was the winner for everything below 397B with my system prompt, till Qwen3.6 27B changed the game.

What can I say? that 27B model is too good, I cannot deny, but look at my poor Arc 770, which seems be abandoned, I decide to continue the journey with my little 'old' Queen,

So here the model is.

For all GPU poor.

Inference: ik_llama.cpp

Thanks:

Qwen Team for bring the base model;

Kimi as assistant;

llama.cpp to make local inference possible;

ik_llama.cpp for SOTA Quants, and faster inference for cpu.