Modelpublic

minjaechoi/qwen3next-80b-a3b-2p12bit-r26

sourceHugging Faceupdated 6d agoView on Hugging Face

0likes295downloads

Model Card

Qwen3-Next-80B-A3B-Thinking — 2.125-bit routed experts (r26)

Internal research checkpoint. Routed experts average 2.125 bits; every other weight is BF16. Weights are stored dequantized in BF16 tensors and load with stock transformers / vLLM.


Base model	Qwen/Qwen3-Next-80B-A3B-Thinking
Routed-expert average	2.125 bits
Internal ID	r26

License follows the base model.