https://huggingface.co/nightmedia/Qwen3.5-9B-Brainwaves-OneJev1.6
#3023
by nightmedia - opened
Dear Team Radermacher,
Could you please quant this model?
https://huggingface.co/nightmedia/Qwen3.5-9B-Brainwaves-OneJev1.6
arc arc/e boolq hswag obkqa piqa wino
bf16 0.700,0.878,0.906,0.775,0.520,0.800,0.728
q8-hi 0.703,0.880,0.906,0.774,0.524,0.802,0.730
q6-hi 0.704,0.880,0.906,0.772,0.530,0.803,0.725
q4-hi 0.701,0.880,0.897,0.776,0.518,0.802,0.717
Quant Perplexity Peak Memory Tokens/sec
q8-hi 4.105 ± 0.026 16.86 GB 702
q6-hi 4.108 ± 0.026 14.62 GB 641
gsm8k
q8-hi 90.07%
q6-hi 90.52%
q4-hi 89.54%
eventually if you can, this one too:
https://huggingface.co/nightmedia/Qwen3.8-27B-MindMeld-OneJev1.6
arc arc/e boolq hswag obkqa piqa wino
bf16 0.747,0.901,0.915
mxfp8 0.744,0.902,0.915,0.841,0.534,0.837,0.801
q8-hi 0.747,0.902,0.915,0.841,0.530,0.832,0.801
q6-hi 0.748,0.899,0.916,0.840,0.534,0.835,0.797
q5-hi 0.744,0.899,0.915,0.840,0.542,0.836,0.795
q4-hi 0.747,0.903,0.916,0.844,0.538,0.838,0.803
mxfp4 0.742,0.897,0.911,0.831,0.524,0.832,0.804
Quant Perplexity Peak Memory Tokens/sec
mxfp8 3.791 ± 0.023 34.74 GB 211
q8-hi 3.779 ± 0.023 37.26 GB 204
q6-hi 4.018 ± 0.025 30.54 GB 199
q5-hi 3.774 ± 0.023 27.18 GB 200
q4-hi 3.813 ± 0.023 23.82 GB 207
mxfp4 3.857 ± 0.024 21.30 GB 211