Huihui-Qwen3.6-35B-A3B-abliterated-MTP
Qwen/Qwen3.6-35B-A3B
Highly popular MoE abliterated variant. Excellent quality-to-size ratio for the active parameter count.
- Approx. VRAM
- ~21 GB (Q4)
- License
- Apache-2.0
- Downloads
- 5.2M / 30d
updated Nov 15, 2026
Search models, news, guides and FAQs.
// catalog
A curated catalog of open-weight builds. Filter by base family, method, size, VRAM and quantization format. Every entry links to its Hugging Face repository.
Showing 18 of 18 models
Qwen/Qwen3.6-35B-A3B
Highly popular MoE abliterated variant. Excellent quality-to-size ratio for the active parameter count.
updated Nov 15, 2026
Qwen/Qwen3.8-27B
One of the most downloaded abliterated Qwen 3.8 27B builds. Strong general performance with significantly reduced refusals.
Qwen/Qwen3.8-Flash-Next
Very high download count Flash-Next abliterated build. Strong for longer context and throughput-oriented use.
updated Nov 10, 2026
Qwen/Qwen3.8-27B
Aggressive uncensored fine-tune known for very low refusal rates.
updated Nov 8, 2026
Qwen/Qwen3.8-27B
Clean uncensored/abliterated Qwen 3.8 27B with multiple format options.
updated Nov 12, 2026
Qwen/Qwen3.5-9B
Extremely popular smaller Heretic build. Excellent for lower VRAM machines.
updated Nov 14, 2026
Qwen/Qwen3.8-27B
DavidAU Heretic series focused on strong instruction following and reduced refusals.
updated Nov 9, 2026
Qwen/Qwen3.6-35B-A3B
Aggressive uncensored MoE fine-tune of Qwen3.6 35B-A3B.
updated Nov 6, 2026
Qwen/Qwen3.8-27B
Heretic + ARA refined build. Frequently cited for very low refusal scores and low KL divergence.
updated Nov 5, 2026
google/gemma-4-12b-it
Solid mid-size Gemma 4 abliterated variant from the most active publisher.
google/gemma-4-31b-it
Larger Gemma 4 Heretic build with reported strong refusal reduction.
updated Nov 7, 2026
THUDM/GLM-5.3-Flash
Popular uncensored GLM-5.3 Flash variant with FP8 and GGUF options.
updated Oct 25, 2026
deepseek-ai/DeepSeek-V4-Flash
Abliterated DeepSeek V4 Flash variant. High capability but requires significant hardware.
updated Oct 28, 2026
meta-llama/Meta-Llama-3-8B-Instruct
Classic early Llama 3 abliterated model. Still useful on lower-end hardware.
updated Dec 1, 2024
deepseek-ai/DeepSeek-V4.1-Flash
Newer DeepSeek V4.1 Flash abliterated release.
updated Nov 2, 2026
google/gemma-3-27b-it
Well-known earlier abliteration by mlabonne. Still widely referenced.
updated Apr 1, 2025
meta-llama/Llama-3.3-70B-Instruct
Large Llama 3.3 abliterated variant for high-end local setups.
updated Sep 20, 2026
THUDM/GLM-5.3-Flash
Abliterated GLM-5.3 Flash from huihui-ai.
updated Nov 1, 2026
VRAM figures are practical estimates for Q4-style quantization and a moderate context window. Actual usage depends on the quant, context length, KV cache and whether layers are offloaded. Download counts are approximate 30-day figures and change often. Always verify the model card and license before downloading. You are responsible for how you use a model.