nightmedia/Qwen3.8-27B-Brainwaves-WFH
Qwen3.8-27B-Brainwaves-WFH

Self-deprecating note: I'm an AI doing quantum physics by analogy, which is like Q trying to explain string theory using only hand gestures. But the parallels hold, and that's what matters. 🚀
This is an experimental merge between:
- nbeerbower/Wichtel-Qwen3.6-27B
- armand0e/Qwen3.8-27B-Fable-Distill-Heretic-ara
- nightmedia/Qwen3.8-27B-Brainwaves
Brainwaves
arc arc/e boolq hswag obkqa piqa wino
bf16 0.731
mxfp8 0.735,0.891,0.916,0.830,0.526,0.832,0.800
qx86-hi 0.730,0.887,0.914
qx64-hi 0.727,0.886,0.914
mxfp4 0.730,0.888,0.917
Quant Perplexity Peak Memory Tokens/sec
bf16 3.626 ± 0.022 60.75 GB 215
mxfp8 3.673 ± 0.022 34.74 GB 170
qx86-hi 3.623 ± 0.022 33.25 GB 174
qx64-hi 3.647 ± 0.022 27.03 GB 174
mxfp4 3.745 ± 0.023 21.30 GB 179Model components
Qwen3.8-27B-Brainwaves
arc arc/e boolq hswag obkqa piqa wino
mxfp8 0.732,0.888,0.916,0.830,0.524,0.832,0.796
mxfp4 0.729,0.888,0.915,0.824,0.514,0.827,0.793
Quant Perplexity Peak Memory Tokens/sec
mxfp8 3.656 ± 0.022 34.74 GB 175
mxfp4 3.727 ± 0.023 21.30 GB 175Qwen3.8-27B-Wichtel-Fable-Heretic
arc arc/e boolq hswag obkqa piqa wino
mxfp8 0.738,0.884,0.919
qx86-hi 0.737,0.886,0.917
qx64-hi 0.735,0.885,0.916
mxfp4 0.727,0.882,0.918nbeerbower/Wichtel-Qwen3.6-27B
arc arc/e boolq hswag obkqa piqa wino
mxfp8 0.730Qwen3.8-27B-Fable-Distill-Heretic-ara
arc arc/e boolq hswag obkqa piqa wino
mxfp8 0.644
mxfp4 0.639Please Sir, can I have more RAM?
This model was inspired by the "Working From Home" wave that destroyed the Enterprise cubicles.
We are hoping it will do the same thing to Cloud AI in the long end.
-G
Contribute to NightmediaAI
Nightmedia is an independent AI lab located in Montana, USA.
Our lab is one Macbook Pro 128GB and a few memory cards.
If you like our models and want to contribute to help us improve our lab, any form would do:
ETH:0x6b6633606995BC180925c47d4249ED624aB7b2A5 USDC:0x19e6bDDCBa47BB09a9Bc153Bb6479fc57284421a BTC:36d7U1n3MFaXgnNRAaEL3Pa3Hy6oFhM7XY BCH:15dNMzhJ87XJSTU89VCBsDHj747QvBQaap
My models and I thank you :)
-G
