463 likesNew · 25d old
Abliterated Qwen3.8 Model Swaps 144 Tensors for Uncensored Run
What is Qwen3.8-Flash-Next-GSQ-RCO-abliterated-GGUF?
This GGUF model is a quantized version of Qwen3.8-Flash-Next that keeps the original GSQ-RCO compression but replaces 144 internal tensors with pre-computed “abliterated” weights. The swap removes the built-in refusal direction while preserving the model’s size and speed, letting you run the uncensored model locally with low memory.
Who it is for
If you need a locally runnable Qwen3.8 model without safety filters and with efficient quantization, this build lets you experiment without the original refusal mechanism.