Rawdog sparse upcycle (8 experts, 1 shared) of Qwen/QwQ-32B. Note that it'll need training to do anything beyond what baseline QwQ can do at much higher compute requirements.

Unless you want to finetune this, it is not what you're looking for.

Unless you want to finetune this, it is not what you're looking for.

Unless you want to finetune this, it is not what you're looking for.

Downloads last month
0
Safetensors
Model size
250B params
Tensor type
BF16
·
Inference Providers NEW
This model is not currently available via any of the supported Inference Providers.

Model tree for rAIfle/QwQonsortium-8x32B-RAW

Base model

Qwen/Qwen2.5-32B
Finetuned
Qwen/QwQ-32B
Finetuned
(11)
this model