hipfire-models
/

qwen3.6-35b-a3b

Mixture of Experts

mixture-of-experts

Model card Files Files and versions

qwen3.6-35b-a3b

139 GB

Ctrl+K

Ctrl+K

1 contributor

History: 26 commits

schuttdev's picture

drop lobotomized mq4 (Q8-attn splice corrupt; registry now ships mq4r/mq4p)

605b30e verified 3 months ago

.gitattributes

2.08 kB
Add mq4r: plain uniform-MQ4 fast tier (was the mq4 content) 4 months ago
KLD-AWQ-instructions.md

8.55 kB
Upload KLD-AWQ-instructions.md with huggingface_hub 4 months ago
README.md

14.5 kB
Upload README.md with huggingface_hub 4 months ago
qwen3.6-35b-a3b.mfp4

20.2 GB
xet

Upload qwen3.6-35b-a3b.mfp4 with huggingface_hub 4 months ago
qwen3.6-35b-a3b.mq2

11.6 GB
xet

Upload qwen3.6-35b-a3b.mq2 with huggingface_hub 4 months ago
qwen3.6-35b-a3b.mq3p

17.2 GB
xet

Upload qwen3.6-35b-a3b.mq3p with huggingface_hub 4 months ago
qwen3.6-35b-a3b.mq4p

19.8 GB
xet

Upload qwen3.6-35b-a3b.mq4p with huggingface_hub 4 months ago
qwen3.6-35b-a3b.mq4r

18.7 GB
xet

Add mq4r: plain uniform-MQ4 fast tier (was the mq4 content) 4 months ago
qwen3.6-35b-a3b.mq5

23.7 GB
xet

Upload qwen3.6-35b-a3b.mq5 with huggingface_hub 4 months ago
qwen3.6-35b-a3b.mq6

27.7 GB
xet

Upload qwen3.6-35b-a3b.mq6 with huggingface_hub 4 months ago
qwen3.6-35b-a3b.mtp

467 MB
xet

a3b MTP: replace full-vocab head with compressed-vocab (cvs16384) head — +12-18% tok/s, lossless full-vocab accept 3 months ago