Your favorite fastest abliterated/safety removed 3.6 and 3.8 27b?
Summary
A Reddit user asks for recommendations on the fastest abliterated variants of Qwen 3.6 and 3.8 27B models for low-thinking tasks, comparing them and sharing their hardware setup.
Similar Articles
8 uncensored Qwen 3.8 27B variants, one base, 167 GPU hours - Abliterlitics
This article presents a comprehensive benchmark of 8 abliterated variants of the Qwen 3.8 27B model against the base model, using weight analysis, KL divergence, 13 benchmarks, and HarmBench refusal tests over 167 GPU hours. The analysis reveals that surgical edits significantly outperform heavy modifications, with aggressive abliteration causing thinking loops in up to 45% of adversarial responses and chat template manipulation detected in some variants.
What is the best Qwen3.8 27b Abliterated version out there?
A user inquires about the best uncensored version of the Qwen3.8 27b AI model for reverse engineering legacy code to avoid refusals.
Benchmark results: what is the best and fastest engine to run Qwen3.8-27B on macOS
This article benchmarks various engines for running the Qwen 3.8 27B model on macOS, comparing their speed and performance in agentic coding tasks, and recommends MTPLX or llama.cpp with MTP for best results.
85 GPU-hours comparing 5 abliteration methods on Qwen3.6-27B: benchmarks, safety, weight forensics - Abliterlitics
This post presents Abliterlitics, an open-source toolkit for analyzing abliteration techniques, and compares five abliteration variants of Qwen3.6-27B using 85 GPU-hours of benchmarks, safety evaluations, and weight forensics. Heretic and Huihui show best capability preservation while all achieve near-complete safety removal.
Nemotron 3.5 Bolt vs Nemo 3.5 Super vs Qwen 3.8 27B/FN
A side-by-side benchmark comparison of Nemotron 3.5 Bolt, Nemotron 3.5 Super, and Qwen 3.8 (27B / FN) across tasks like SWE-bench Pro, Terminal Bench, GPQA-Diamond, and HLE, showing Qwen 3.8 FN leading on most metrics.