mradermacher

/

model_requests

like 176

Dear Team Radermacher,
It's been a while--if it's not too much squabble, I have a Holo3 new Qwable with a Qwopus:

https://huggingface.co/nightmedia/Qwen3.6-35B-A3B-Qwable-Holo3-Qwopus

This is a NuSLERP merge of:

• samuelcardillo/Qwopus-MoE-35B-A3B
• Hcompany/Holo3-35B-A3B
• lordx64/Qwable-v1

         arc   arc/e boolq hswag obkqa piqa  wino
bf16     0.614,0.798,0.894,0.768,0.450,0.825,0.733
mxfp8    0.608,0.788,0.899,0.769,0.454,0.831,0.720
qx86-hi  0.610,0.796,0.896,0.768,0.452,0.826,0.735
qx64-hi  0.618,0.796,0.896,0.767,0.448,0.817,0.719
mxfp4    0.613,0.807,0.885,0.764,0.460,0.826,0.705

Quant    Perplexity      Peak Memory   Tokens/sec
qx64-hi  4.118 ± 0.026   36.91 GB      1499
mxfp4    4.406 ± 0.028   25.33 GB      1414

It does great on MLX, so I was thinking, if you would be so kind and quant this :)

Much appreciated!

-G