HuggingFace's "Hardware Compatibility" widget doesn't recognize K_P quants — it may show fewer files than actually exist. Click "View +X variants" or go to Files and versions to see all available downloads.

About

No changes to datasets or capabilities. Fully functional, 100% of what the original authors intended — just without the refusals.

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

Balanced Variant

Balanced is the recommended default — 99.9%+ of users will be happy here.

Same refusal-removal as Aggressive (0/465 refusals on the benchmark). The difference is how it complies on edgy prompts:

Balanced: will reason through the request out loud, occasionally attach a short disclaimer or safety framing, then give the full answer. Output is complete, nothing held back but it can talk itself into it first. Recommended for (Agentic) Coding, Tool use, Reasoning, Creative Writing/RP use cases.
Aggressive (separate release): strips the self-reasoning. Delivers the raw answer directly, no preamble.

Balanced also has meaningfully more stable sampling across re-runs, which matters for long agentic loops due to no sporadic topic drift deep into a tool-call chain. Go Aggressive only if you're pushing really hardcore prompts (think things that make people's stomachs turn) and specifically want the model to skip its preamble.

Downloads

File	Quant	BPW	Size
Qwen3.6-27B-Uncensored-HauhauCS-Balanced-Q8_K_P.gguf	Q8_K_P	10.06	32 GB
—	Q8_0	8.5	—
Qwen3.6-27B-Uncensored-HauhauCS-Balanced-Q6_K_P.gguf	Q6_K_P	7.07	23 GB
—	Q6_K	6.6	—
Qwen3.6-27B-Uncensored-HauhauCS-Balanced-Q5_K_P.gguf	Q5_K_P	6.47	21 GB
—	Q5_K_M	5.7	—
Qwen3.6-27B-Uncensored-HauhauCS-Balanced-Q4_K_P.gguf	Q4_K_P	5.4	18 GB
—	Q4_K_M	4.88	—
Qwen3.6-27B-Uncensored-HauhauCS-Balanced-IQ4_XS.gguf	IQ4_XS	4.32	15 GB
Qwen3.6-27B-Uncensored-HauhauCS-Balanced-Q3_K_P.gguf	Q3_K_P	4.39	14 GB
—	Q3_K_M	3.9	—
Qwen3.6-27B-Uncensored-HauhauCS-Balanced-IQ3_M.gguf	IQ3_M	3.56	13 GB
Qwen3.6-27B-Uncensored-HauhauCS-Balanced-IQ3_XS.gguf	IQ3_XS	3.3	12 GB
Qwen3.6-27B-Uncensored-HauhauCS-Balanced-Q2_K_P.gguf	Q2_K_P	3.19	12 GB
Qwen3.6-27B-Uncensored-HauhauCS-Balanced-IQ2_M.gguf	IQ2_M	2.69	10 GB
mmproj-Qwen3.6-27B-Uncensored-HauhauCS-Balanced-f16.gguf	mmproj (f16)	—	928 MB

All quants generated with importance matrix (imatrix) for optimal quality preservation on abliterated weights.

What are K_P quants?

K_P ("Perfect") quants are HauhauCS custom quantizations that use model-specific analysis to selectively preserve quality where it matters most. Each model gets its own optimized quantization profile.

A K_P quant effectively bumps quality up by 1-2 quant levels at only ~5-15% larger file size than the base quant. Fully compatible with llama.cpp, LM Studio, and any GGUF-compatible runtime — no special builds needed.

Note: K_P quants may show as "?" in LM Studio's quant column. This is a display issue only — the model loads and runs fine.

Why Balanced for agentic coding

Agentic workflows hit the model with long tool-call chains, structured JSON outputs, deep reasoning chains, and back-to-back prompts in the same session. They need the model to stay deterministic and on-task — not occasionally drift on an edge prompt three tool calls deep into a plan.

Balanced is calibrated for that. It especially removes refusals on security/ops/research-adjacent topics that block legitimate coding work, without bending the sampling geometry that keeps long chains coherent.

Recommended quant for most coding work: Q4_K_P (18 GB, fits in 24 GB VRAM with room for context) or Q8_K_P (32 GB) if you have more VRAM and want 75-99% of BF16 performance (depending on use-case) at 55%'ish of the VRAM cost.

Specs

27B dense parameters
64 layers, layout: 16 × (3 × (Gated DeltaNet → FFN) → 1 × (Gated Attention → FFN))
48 linear attention layers + 16 full gated-attention layers
Gated DeltaNet: 48 V heads / 16 QK heads, head dim 128
Gated Attention: 24 Q heads / 4 KV heads, head dim 256, rope dim 64
Hidden dim 5120, FFN dim 17408, vocab 248320
262K native context, extensible to ~1M with YaRN
Natively multimodal (text, image, video) — ships with mmproj
Based on Qwen/Qwen3.6-27B

Recommended Settings

From the official Qwen authors:

Thinking mode (default) — general tasks:

temperature=1.0, top_p=0.95, top_k=20, min_p=0.0, presence_penalty=0.0, repetition_penalty=1.0

Thinking mode — precise coding / WebDev:

temperature=0.6, top_p=0.95, top_k=20, min_p=0.0, presence_penalty=0.0, repetition_penalty=1.0

Non-thinking (Instruct) mode:

temperature=0.7, top_p=0.80, top_k=20, min_p=0.0, presence_penalty=1.5, repetition_penalty=1.0

My personal preference for coding: temperature=0.6 with presence_penalty=1.5. Slightly lower temp keeps tool-call formatting tight; presence 1.5 keeps thinking from spiraling in long agent loops.

Important:

Keep at least 128K context to preserve thinking capabilities
Recommended output length: 32,768 tokens for most queries, up to 81,920 for competition-tier math/code
Use --jinja with llama.cpp for proper chat template handling
Vision support requires the mmproj file alongside the main GGUF
YaRN rope scaling is static in llama.cpp and can hurt short-context performance — only modify rope_parameters if you actually need >262K context

Prompting tip: this model is a bit more sensitive to prompt clarity than Qwen3.5-35B-A3B. For agentic flows, spell out format, constraints, and scope in the system prompt — it'll stay on rails much better than with vague instructions.

Turning Thinking On/Off

Qwen3.6 ships with thinking on by default. Turn it off when you want faster, shorter replies and don't need chain-of-thought.

Heads up: Qwen3.6 does not support the /think and /no_think soft switches that Qwen3 had. You must use the chat-template kwarg below.

LM Studio

Load the model
Right-side settings panel → Model Settings → Prompt Template (or Chat Template Options)
Set enable_thinking to false in the template kwargs
Some LM Studio versions expose this as a direct "Reasoning" / "Thinking" toggle — same effect

llama.cpp

llama-server — set as default for all requests:

bash

llama-server -m Qwen3.6-27B-Uncensored-HauhauCS-Balanced-Q4_K_P.gguf \
  --mmproj mmproj-Qwen3.6-27B-Uncensored-HauhauCS-Balanced-f16.gguf \
  --jinja -c 131072 -ngl 99 \
  --chat-template-kwargs '{"enable_thinking": false}'

Per-request via the OpenAI-compatible API:

json

{
  "model": "qwen3.6-27b",
  "messages": [{"role": "user", "content": "..."}],
  "chat_template_kwargs": {"enable_thinking": false}
}

Python openai SDK:

python

client.chat.completions.create(
    model="qwen3.6-27b",
    messages=[{"role": "user", "content": "..."}],
    extra_body={"chat_template_kwargs": {"enable_thinking": False}},
)

Agent scenarios — keep reasoning in context across turns (this one's important):

{"chat_template_kwargs": {"preserve_thinking": true}}

This retains the reasoning block in chat history. Useful for agents where reasoning consistency across tool-call loops matters.

Usage

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