Qwen3.5-abliterated-MAX-AIO-GGUF

Qwen3.5-abliterated-MAX-AIO-GGUF

prithivMLmods/Qwen3.5 Unredacted-MAX models (0.8B, 2B, 4B, 9B, 27B variants) are experimental "abliterated" evolution of Alibaba's official Qwen3.5 series, applying advanced refusal-minimization techniques to remove safety alignments and internal censorship patterns while preserving or enhancing core reasoning, coding, math, and multimodal capabilities across scales. Created by prithivMLmods as part of their "Unredacted MAX" collection—similar to their Qwen3-VL abliterated releases like Qwen3-VL-8B-Instruct-Unredacted-MAX—these models undergo continual "abliteration" training for unrestricted instruction adherence, dense output generation, and reduced hallucinations on diverse prompts without standard safety refusals, making them suitable for edge deployment (FP8/GGUF quants available) in research, dataset creation, or applications needing uncensored reasoning. Scaling from ultra-light 0.8B (mobile/Pi) to powerful 27B, they retain Qwen3.5's hybrid Gated DeltaNet architecture, 262K context, 201-language vocab, and toggleable thinking modes while prioritizing raw capability over guardrails, outperforming vanilla counterparts in open-ended tasks per community benchmarks though lacking official evaluations.

<div style=" background: rgba(61, 122, 255, 0.15); padding: 16px; border-radius: 6px; border: 1px solid rgba(0, 102, 255, 0.3); margin: 16px 0; "> <details> <summary>Open this to view the entire directory.</summary>

prithivMLmods/Qwen3.5-abliterated-MAX-AIO-GGUF (main)
+-- README.md (9.9 KB)
+-- .gitattributes (4.8 KB)
+-- config.json (32 B)
+-- Qwen3.5-0.8B-Unredacted-MAX-GGUF
|   +-- Qwen3.5-0.8B-Unredacted-MAX.F32.gguf (2.8 GB)
|   +-- Qwen3.5-0.8B-Unredacted-MAX.BF16.gguf (1.4 GB)
|   +-- Qwen3.5-0.8B-Unredacted-MAX.F16.gguf (1.4 GB)
|   +-- Qwen3.5-0.8B-Unredacted-MAX.Q8_0.gguf (774.2 MB)
|   +-- Qwen3.5-0.8B-Unredacted-MAX.mmproj-f32.gguf (383.7 MB)
|   +-- Qwen3.5-0.8B-Unredacted-MAX.mmproj-bf16.gguf (197.7 MB)
|   +-- Qwen3.5-0.8B-Unredacted-MAX.mmproj-f16.gguf (197.7 MB)
|   +-- Qwen3.5-0.8B-Unredacted-MAX.mmproj-q8_0.gguf (110.6 MB)
+-- Qwen3.5-2B-Unredacted-MAX-GGUF
|   +-- Qwen3.5-2B-Unredacted-MAX.F32.gguf (7.0 GB)
|   +-- Qwen3.5-2B-Unredacted-MAX.BF16.gguf (3.5 GB)
|   +-- Qwen3.5-2B-Unredacted-MAX.F16.gguf (3.5 GB)
|   +-- Qwen3.5-2B-Unredacted-MAX.Q8_0.gguf (1.9 GB)
|   +-- Qwen3.5-2B-Unredacted-MAX.mmproj-f32.gguf (1.2 GB)
|   +-- Qwen3.5-2B-Unredacted-MAX.mmproj-bf16.gguf (640.3 MB)
|   +-- Qwen3.5-2B-Unredacted-MAX.mmproj-f16.gguf (640.3 MB)
|   +-- Qwen3.5-2B-Unredacted-MAX.mmproj-q8_0.gguf (347.8 MB)
+-- Qwen3.5-4B-Unredacted-MAX-GGUF
|   +-- Qwen3.5-4B-Unredacted-MAX.F32.gguf (15.7 GB)
|   +-- Qwen3.5-4B-Unredacted-MAX.BF16.gguf (7.8 GB)
|   +-- Qwen3.5-4B-Unredacted-MAX.F16.gguf (7.8 GB)
|   +-- Qwen3.5-4B-Unredacted-MAX.Q8_0.gguf (4.2 GB)
|   +-- Qwen3.5-4B-Unredacted-MAX.mmproj-f32.gguf (1.2 GB)
|   +-- Qwen3.5-4B-Unredacted-MAX.mmproj-bf16.gguf (644.3 MB)
|   +-- Qwen3.5-4B-Unredacted-MAX.mmproj-f16.gguf (644.3 MB)
|   +-- Qwen3.5-4B-Unredacted-MAX.mmproj-q8_0.gguf (349.9 MB)
+-- Qwen3.5-9B-Unredacted-MAX-GGUF
    +-- Qwen3.5-9B-Unredacted-MAX.F32.gguf (33.4 GB)
    +-- Qwen3.5-9B-Unredacted-MAX.BF16.gguf (16.7 GB)
    +-- Qwen3.5-9B-Unredacted-MAX.F16.gguf (16.7 GB)
    +-- Qwen3.5-9B-Unredacted-MAX.Q8_0.gguf (8.9 GB)
    +-- Qwen3.5-9B-Unredacted-MAX.mmproj-f32.gguf (1.7 GB)
    +-- Qwen3.5-9B-Unredacted-MAX.mmproj-bf16.gguf (879.0 MB)
    +-- Qwen3.5-9B-Unredacted-MAX.mmproj-f16.gguf (879.0 MB)
    +-- Qwen3.5-9B-Unredacted-MAX.mmproj-q8_0.gguf (595.3 MB)

</details> </div>

Model List

Qwen3.5-0.8B-Unredacted-MAX

Qwen3.5-2B-Unredacted-MAX

Qwen3.5-4B-Unredacted-MAX

Qwen3.5-9B-Unredacted-MAX

Quants Usage

(sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)

Here is a handy graph by ikawrakow comparing some lower-quality quant types (lower is better):