Uncensored / Abliterated GGUF Models - a 0bserverx Collection

0bserverx 's Collections

updated Sep 1

GGUF quants of abliterated (uncensored) models for local llama.cpp inference. Curated by 0bserverx.

0bserverx/Muse-Glimmer-30B-Heretic-Uncensored-GGUF

Image-Text-to-Text • 28B • Updated Aug 20 • 22.3k • 30

Note Abliterated GGUF of Muse Glimmer 30B (multimodal, vision built-in). Q4_K_S 16GB, 38 t/s on RTX 4080, zero refusals. Base: meta-models/Muse-Glimmer-30B.
0bserverx/Qwen3.8-27B-Heretic-Abliterated-Uncensored-GGUF

Text Generation • 27B • Updated Sep 20 • 1.15M • 818

Note RVN — double-refined ARA, refusals 0-1/100, KL 0.0085
0bserverx/RVN-Qwen3.8-Flash-Next-Abliterated-Uncensored

Text Generation • 177B • Updated Sep 1 • 650 • 12

Note RVN V6 main/text checkpoint from the official Qwen3.8-Flash-Next base. Discloses RR100 behavior, protected-gate results, KL boundaries, matched Q6_K throughput/perplexity, and downstream lm-eval comparisons.
0bserverx/RVN-Qwen3.8-Flash-Next-Abliterated-Uncensored-GGUF

Text Generation • 177B • Updated Sep 1 • 23.1k • 34

Note RVN V6 GGUF release: 18 verified quant families across 144 shards, including five importance-matrix families, with matched Q6_K runtime, perplexity, and downstream lm-eval results.