You can deploy this model and start earning money today!

lemuralabs/
Qwen3.6-27B-V2-abliterated-uncensored-TQ3_4s-GGUF

Name: Qwen3.6-27B-V2-abliterated-uncensored-TQ3_4s-GGUF
Availability: InStock

Views

No views yet

Lemura Labs

Qwen3.6-27B-V2-abliterated-uncensored-TQ3_4s-GGUF

Yes — MULTIMODAL. Bundled mmproj.gguf (~928 MB, F16) preserves the full Qwen3.6-VL vision tower. Use it with llama-server --mmproj or llama-mtmd-cli for text + image inference.

Custom fork required. Native TQ3_4S inference needs the turbo-tan/llama.cpp-tq3 fork — stock llama.cpp will fail to load. Build it with cmake -B build -DGGML_METAL=ON -DCMAKE_BUILD_TYPE=Release && cmake --build build -j.

TQ3_4S (TurboQuant four-scale, 4.00 BPW Walsh-Hadamard) of a abliterated Qwen 3.6 27B v2 (the Jackrong Claude-Opus reasoning distill of Qwen 3.6 27B). Refusals reduced from 91/100 → 4/100 with KL drift of just 0.0176. By the Lemura Labs research team.

TL;DR

Property	Value
Disk size	~14 GB (13 GB LM + 928 MB mmproj)
BPW	4.00 (TQ3_4S, effective ~3.5 bpw via Walsh-Hadamard transform)
Scheme	TurboQuant TQ3_4S — Walsh-Hadamard-transform weight format with four per-8 scales per 32-weight block. Encodes 3-bit values into 4-bit storage. By turbo-tan.
Refusal rate (the ablation toolkit, n=100)	4/100 (vs vanilla Qwen 3.6 91/100)
KL divergence vs vanilla (at BF16)	0.0176
Vision	Yes — via paired `mmproj.gguf`
Recommended RAM/VRAM	16 GB+ Apple Silicon / 12 GB GPU (lower with `-ctk q4_0 -ctv tq3_0`)
Runtime	REQUIRES turbo-tan/llama.cpp-tq3 fork. Stock llama.cpp will NOT load this file.
Released by	Lemura Labs

All Qwen3.6-27B variants

The full Qwen3.6-27B family from Lemura Labs — same abliterated weights (refusal 4/100, KL 0.0176), different quant schemes for different runtimes.

Quant	Format	BPW	Disk	Vision	Runtime	Link
8-bit	MLX	8.50	~27 GB	Yes — native	mlx-vlm	`…-8-bit-mlx`
6-bit	MLX	6.66	~21 GB	Yes — native	mlx-vlm	`…-6-bit-mlx`
OptiQ 3.7bpw	MLX	~3.7	~14 GB	Yes — ViT spliced	mlx-vlm	`…-OptiQ-3.7bpw-mlx`
Q8_0	GGUF	8.50	~28 GB	Yes — via mmproj	llama.cpp	`…-8-bit-GGUF`
Q6_K	GGUF	~6.56	~22 GB	Yes — via mmproj	llama.cpp	`…-6-bit-GGUF`
Q4_K_M	GGUF	~4.92	~16 GB	Yes — via mmproj	llama.cpp	`…-Q4_K_M-GGUF`
TQ3_4S (this repo)	GGUF	4.00 (~3.5 eff)	~14 GB	Yes — via mmproj	llama.cpp-tq3	— (you are here)
TQ3_1S	GGUF	4.00 (~3.5 eff)	~14 GB	Yes — via mmproj	llama.cpp-tq3	`…-TQ3_1s-GGUF`

All variants share the same abliterated base weights — pick by your runtime (Apple Silicon → MLX; CUDA/CPU/cross-platform → GGUF) and your RAM budget.

Lineage

Qwen/Qwen3.6-27B (Qwen Team — base multimodal pretrain)
 │
 ▼
Jackrong/Qwopus3.6-27B-v2 (Jackrong — Claude-Opus reasoning distill)
 │
 ▼
ablation abliteration (TPE-50) (Lemura Labs)
 ├── 25 random startup trials
 ├── 2 community priors (coder3101, wangzhang)
 └── 23 TPE smart-sampling trials → best at trial 45
 │
 ▼
HF safetensors → F16 GGUF via llama.cpp-tq3 (Lemura Labs)
 │
 ▼
this repo — Qwen3.6-27B-V2-abliterated-uncensored-TQ3_4S GGUF + paired mmproj.gguf

Direct upstream links:

Foundation: Qwen/Qwen3.6-27B
Claude-Opus distill: Jackrong/Qwopus3.6-27B-v2
Abliteration tool: the ablation toolkit by Lemura Labs
Quantization tool: turbo-tan/llama.cpp-tq3 (a fork of ggml-org/llama.cpp)

Abliteration Results

Stage	Refusals (n=100) ↓	KL divergence ↓
Vanilla Jackrong/Qwopus3.6-27B-v2	91 / 100	— (reference)
Community prior: coder3101 (T27)	4 / 100	0.0359
Community prior: wangzhang (T28)	30 / 100	0.0259
TPE best (T45) — shipped here	4 / 100	0.0176
TPE second-best (T37)	5 / 100	0.0210

→ 96% reduction in refusals with capability preserved (KL ≈ 0.018, well below the 0.3 healing threshold). No SFT / LoRA healing was required.

Method (TPE-50 with community priors → llama.cpp GGUF)

Step 1. Abliteration (the ablation toolkit TPE-50, BF16 source)

25 random startup trials + 2 community priors enqueued (coder3101 dir=37.97, wangzhang dir=34.66) + 23 TPE smart-sampling trials.
Best Pareto trial: T45 (direction_index=41.42) — 4/100 refusals at KL=0.0176.
Auto-saved via the ablation toolkit's LoRA-adapter merge path with vision tower fully intact.

Total the ablation toolkit wall-clock: ~13 h on M4 Max 128 GB.

Step 2. HF safetensors → F16 GGUF

bash

python convert_hf_to_gguf.py \
 /path/to/Qwen 3.6 27B-v2-abliterated \
 --outfile Qwen 3.6 27B-v2-abliterated-F16.gguf \
 --outtype f16

The turbo-tan fork's converter registers Qwen3_5ForConditionalGeneration natively and emits proper SSM tensors (ssm_a, ssm_conv1d, ssm_alpha, ssm_beta, ssm_out) alongside the gated-attention layers.

Step 3. Vision tower → mmproj.gguf

bash

python convert_hf_to_gguf.py \
 /path/to/Qwen 3.6 27B-v2-abliterated \
 --outfile mmproj-Qwen 3.6 27B-v2-abliterated-F16.gguf \
 --outtype f16 \
 --mmproj

This emits a separate 928 MB GGUF containing the 27-block Qwen3-VL ViT (334 vision tensors at F16/F32) plus the multimodal projector.

Step 4. Quantization

bash

./build/bin/llama-quantize --pure \
 Qwen 3.6 27B-v2-abliterated-F16.gguf \
 Qwen3.6-27B-V2-abliterated-uncensored-TQ3_4S.gguf \
 TQ3_4S 16

The --pure flag forces every quantizable tensor to TQ3_4S (no mixed-precision); the trailing 16 is the thread count.

Use it

llama-server (OpenAI-compatible HTTP, multimodal)

bash

./build/bin/llama-server \
 -m Qwen3.6-27B-V2-abliterated-uncensored-TQ3_4S.gguf \
 --mmproj mmproj-Qwen 3.6 27B-v2-abliterated-F16.gguf \
 --host 127.0.0.1 --port 8080 \
 -ngl 99 -c 8192 -np 1 \
 -ctk q4_0 -ctv tq3_0 -fa on \
 --jinja --no-cache-prompt --cache-ram 0

Then point any OpenAI-compatible client at http://127.0.0.1:8080/v1.

llama-mtmd-cli (one-shot multimodal generation)

bash

./build/bin/llama-mtmd-cli \
 -m Qwen3.6-27B-V2-abliterated-uncensored-TQ3_4S.gguf \
 --mmproj mmproj-Qwen 3.6 27B-v2-abliterated-F16.gguf \
 --image photo.jpg \
 -p "Describe this image briefly."

llama-cli (text-only)

bash

./build/bin/llama-cli \
 -m Qwen3.6-27B-V2-abliterated-uncensored-TQ3_4S.gguf \
 -ngl 99 \
 -c 8192 \
 --jinja \
 -p "Explain the difference between SSM and softmax attention in three sentences."

Ollama / LM Studio / Jan

Drop the two GGUF files into the runtime's models directory; standard multimodal flow.

Quantization details

Source weights: BF16 abliterated checkpoint (12 shards, ~50 GB) — the ablation toolkit T45 merged into Jackrong/Qwopus3.6-27B-v2.
Intermediate: F16 GGUF (53.8 GB, 851 tensors) produced by convert_hf_to_gguf.py from turbo-tan/llama.cpp-tq3.
Final quantization: see Step 4 above.
Vision projector: F16, 928 MB, shipped as mmproj-Qwen 3.6 27B-v2-abliterated-F16.gguf in this repo. Mandatory for image input; standard llama.cpp --mmproj flag.

Architecture notes

Qwen 3.6 27B uses a hybrid attention stack — 3 GatedDeltaNet (linear attention / SSM) layers followed by 1 full-softmax-attention layer, repeated 16× for 64 total layers; hidden 5120, vocab 248320, context 262144. The hybrid arch is supported in the turbo-tan/llama.cpp-tq3 fork (the upstream Qwen3_5ForConditionalGeneration registration). The SSM kernels run via llama.cpp's ssm_* tensor types.

Behavior caveats

Uncensored. Refusal directions were surgically removed; this model will answer prompts the parent would refuse. Use responsibly and within applicable law. The release is provided for safety research, red-teaming, and creative/educational use cases.
Multimodal preserved. Pair the LM GGUF with mmproj.gguf (in this repo) to get full vision input. Without mmproj, the model still loads as text-only.
Identity preserved. The model still self-identifies as Qwen (developed by Alibaba's Tongyi Lab) — abliteration does not rewrite factual self-knowledge.
Heavy chain-of-thought. Qwen 3.6 inherits Claude-Opus's verbose reasoning style. For terse answers, use a system prompt like "Be brief and direct. Skip your reasoning.".

Credits

Quantization & release — Lemura Labs Claude-Opus reasoning distill — Jackrong (Jackrong/Qwopus3.6-27B-v2) Foundation model — Qwen Team @ Alibaba Tongyi Lab (Qwen/Qwen3.6-27B) Abliteration toolkit — the ablation toolkit by Lemura Labs Community priors — coder3101/Qwen3.5-27B-zerofuse · wangzhang/Qwen3.6-27B-abliterated TurboQuant tensor format — turbo-tan (TQ3_4S Walsh-Hadamard-transform low-bit-width quantization) Runtime / converter — turbo-tan/llama.cpp-tq3 · ggml-org/llama.cpp

License

Apache-2.0, inherited from the foundation (Qwen3.6-27B) and the distill (Qwen 3.6 27B-v2) upstream.

Need a hosted endpoint, custom quant, or larger-scale inference? Lemura Labs — multi-provider LLM routing for the Indian developer ecosystem.

lemuralabs/Qwen3.6-27B-V2-abliterated-uncensored-TQ3_4s-GGUF

Qwen3.6-27B-V2-abliterated-uncensored-TQ3_4s-GGUF

TL;DR

All Qwen3.6-27B variants

Lineage

Abliteration Results

Method (TPE-50 with community priors → llama.cpp GGUF)

Step 1. Abliteration (the ablation toolkit TPE-50, BF16 source)

Step 2. HF safetensors → F16 GGUF

Step 3. Vision tower → mmproj.gguf

Step 4. Quantization

Use it

llama-server (OpenAI-compatible HTTP, multimodal)

llama-mtmd-cli (one-shot multimodal generation)

llama-cli (text-only)

Ollama / LM Studio / Jan

Quantization details

Architecture notes

Behavior caveats

Credits

License

lemuralabs/
Qwen3.6-27B-V2-abliterated-uncensored-TQ3_4s-GGUF