This model is not meant to be used as is, it will just say nonsensical things and loop endlessly. I have uploaded this to request a repair run as I do not have the hardware to do it myself. I would greatly appreciate it if anyone is willing to run a repair run on it.

I did try doing it myself but unfortunately I used a contaminated dataset which got it most of the way there but introduced some significant problems, so this is the "pure" pruned version of Qwen3.5-35B-A3B.

Vision is fully intact, but the aggressive pruning makes the model loop and malfunction.

Downloads last month
43
Safetensors
Model size
7B params
Tensor type
BF16
·
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for blascotobasco/Qwen3.5-32E-A3B-Test

Finetuned
(87)
this model