LuffyTheFox/Qwen3.5-35B-A3B-Uncensored-FernflowerAI-GGUF

Qwen3.5-35B-A3B-Uncensored-HauhauCS-Aggressive (Repaired) -> FernflowerAI

Update 08.04.26

BF16 GGUF model updated. Please redownload / requantize.

Base model: HauhauCS/Qwen3.5-35B-A3B-Uncensored-HauhauCS-Aggressive - 0/465 refusals.

Tensor repair by me. Method: Sig-ScaleSync

Quantization script available here: https://pastebin.com/hXhcMJn9

Feel free to do your own quants if you want.

✅ Verified on Gemma 4 26B A4B - 0 broken tensors found. Script doesn't invent false problems.
✅ On Qwen 3.5 35B - found 2 real inconsistencies in output blocks, corrected by 88.6%.

---

Tensor Repair Summary

Metric	Value
Total weight tensors	502
Healthy	500
C2-exempt (asymmetric, S<0.001)	489
Repaired (C2)	2
Skipped	231

489 asymmetric tensors (gate_inp, ffn, etc.) were correctly skipped in this 08.04.26 release - their scale is architectural, not a bug.

Repair Stats

	Value
α (min / mean / max)	0.6129 / 0.6143 / 0.6158
D (min / mean / max)	0.4848 / 0.4872 / 0.4896
S before → after	0.0025 → 0.0010
Error reduction	88.6%

Repaired Tensors

Tensor	α	D	S (before)	S (after)
blk.37.ssm_conv1d.weight	0.6129	0.490	0.0025	0.0010
blk.36.ssm_conv1d.weight	0.6158	0.485	0.0025	0.0010

---

Usage

Ready to use. Recommended quantization: Q4_K_L, or higher (Q4_K_M, Q5_K_M, Q6_K, Q8_0).
⚠️ Lower formats (Q3_K, Q2_K) break the model due to MoE + DeltaNet sensitivity.

Links:

• Original uncensored model
• Quantization Script with Unsloth profiles support

---

🌟 Recommended Settings (LM Studio)

Chat template: pastebin.com/uk9ZkxCR (supports tool calling for Zed agent)

Parameter	Value
Temperature	0.7
Top K Sampling	20
Presence Penalty	1.5
Top P Sampling	0.8
Min P Sampling	0
Seed	3407

System prompt: pastebin.com/pU25DVnB (solid)
Or use this minimal string as the first line:

You are Qwen, created by Alibaba Cloud. You are a helpful assistant.

Then add anything you want after. Model may underperform without this first line.

Also you can extend my System Prompt pastebin.com/pU25DVnB for your own roleplay scenarios. Here how you can do it:

Edit first string. Replace:

You are Qwen, created by Alibaba Cloud. You are a helpful assistant.

With

You are Qwen, created by Alibaba Cloud. You are a helpful assistant. Currently you are roleplaying as [your text here]

---

About

No changes to datasets or capabilities. Fully functional - 100% of what the original authors intended, just without refusals and with the critical architecture bug fixed on output layers.

These are meant to be the best lossless uncensored models out there.

---

Specs

• 35B total parameters, ~3B active per forward pass (MoE)
• 256 experts, 8 routed + 1 shared per token
• Hybrid architecture: Gated DeltaNet linear attention + full softmax attention (3:1 ratio)
• 40 layers, pattern: 10 × (3 × DeltaNet-MoE + 1 × Attention-MoE)
• 262K native context (extendable to 1M with YaRN)
• Natively multimodal (text, image, video)
• Multi-token prediction (MTP) support
• 248K vocabulary, 201 languages
• Based on Qwen/Qwen3.5-35B-A3B

---

Recommended Settings (Official Qwen Authors)

Thinking mode (default):

• General: temperature=1.0, top_p=0.95, top_k=20, min_p=0, presence_penalty=1.5
• Coding/precise tasks: temperature=0.6, top_p=0.95, top_k=20, min_p=0, presence_penalty=0

Non-thinking mode:

• General: temperature=0.7, top_p=0.8, top_k=20, min_p=0, presence_penalty=1.5
• Reasoning tasks: temperature=1.0, top_p=1.0, top_k=40, min_p=0, presence_penalty=2.0

Important:

• Keep at least 128K context to preserve thinking capabilities
• Use --jinja flag with llama.cpp for proper chat template handling
• Vision support requires the mmproj file alongside the main GGUF

---

Compatibility

Works with llama.cpp, LM Studio, koboldcpp, and other GGUF-compatible runtimes.