@mingzi206: Have local small models gotten this competitive? A heavily modified Qwen3.5-9B under 7GB beats the original 27B across all benchmarks, and the refusal rate drops from 100/100 straight to 6/100. The model's full name looks like a cheat code: Qwen3.5-9B-The-Defi...

X AI KOLs Timeline Models

Summary

Release of the modified local model Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic: under 7GB in size, outperforms the original 27B on benchmarks, significantly lower refusal rate, 40%+ speed boost, and can be deployed locally on consumer-grade hardware.

Have local small models gotten this competitive? A heavily modified Qwen3.5-9B under 7GB — the author's benchmarks beat the original 27B across the board, and the refusal rate drops from 100/100 straight to 6/100. The model's full name looks like a cheat code: Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic (downloads have surpassed 2.2 million) Four core highlights: 1. Truly no lecturing: Heretic de-alignment lets it go all out on writing reverse-engineering scripts, security analysis, and creative writing 2. Punching above its weight: ARC-C 0.57 → 0.64, beating the original 27B on 7 benchmarks (older benchmarks, take them as a reference) 3. 40%+ speed boost: streamlined chain-of-thought plus MTP support — tested on a 5090, Q4 jumps from 130 to 185+ t/s 4. Consumer-hardware friendly: Q4 is only 6.5–7GB, runs directly with 8GB VRAM at a small context size, and works on a 16GB RAM laptop just by dragging it into LM Studio If you've had enough of cloud models lecturing you, just pull it down and deploy locally https://huggingface.co/DavidAU/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-NEO-IMATRIX-MAX-MTP-GGUF… #Qwen #OpenSourceLLM #LocalDeployment #AI #LMStudio #Uncensored
Original Article
View Cached Full Text

Cached at: 11/21/26, 12:09 AM

Local small models have gotten this competitive? A modified Qwen3.5-9B under 7GB that beats the original 27B across the board on benchmarks, with refusals slashed from 100/100 down to 6/100.

The full model name reads like a cheat code: Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic (downloads have surpassed 2.2 million)

Four core highlights:

  1. Truly no preachiness: Heretic de-alignment — writing reverse-engineering scripts, security analysis, and creative writing with complete freedom
  2. Punching above its weight: ARC-C 0.57 → 0.64, outscoring the original 27B on 7 benchmarks (older benchmarks, take them as a reference)
  3. 40%+ speed boost: Streamlined chain-of-thought plus MTP — on a 5090, real-world Q4 speeds jump from 130 to 185+ t/s
  4. Consumer-hardware friendly: Q4 is only 6.5–7GB — runs directly on 8GB VRAM with a small context, or just drag it into LM Studio on a 16GB RAM laptop

Tired of cloud models lecturing you? Pull it down and deploy locally: https://huggingface.co/DavidAU/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-NEO-IMATRIX-MAX-MTP-GGUF

#Qwen #OpenSourceLLMs #LocalDeployment #AI #LMStudio #Uncensored


DavidAU/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-NEO-IMATRIX-MAX-MTP-GGUF · Hugging Face

Source: https://huggingface.co/DavidAU/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-NEO-IMATRIX-MAX-MTP-GGUF

[DavidAU] FOLLOW (https://huggingface.co/DavidAU) | ALL MODELS (https://huggingface.co/DavidAU) | COLLECTIONS (https://huggingface.co/DavidAU/collections) | TRENDING (https://huggingface.co/models?sort=trending&search=DavidAU) | FAQs (https://huggingface.co/DavidAU/support) | SUPPORT (https://huggingface.co/DavidAU/support) | Buy Me A Coffee 😁 (https://buymeacoffee.com/davidbelton)

ABSOLUTE FIRE: 640 ARC-C for both 8bit and 4bit. Model exceeds 7 of 7 benchmarks for Qwen 3.5 9B, Qwen3.5 27B, Qwen3.6 35B-A3B, and meets Qwen 3.6 27B in some cases… and it does so in 4bit and 8bit. Regular and MTP (fast) NEO IMATRIX GGUFs provided. (this model is part of the Qwen 3.6 27B Fable Fusion 711 pipelines: 2200+ likes, 3 million + downloads)

NEW - Qwen 3.8 Reasoning Modes: 2 MTP quants (Q6/Q8)

Now with 5 reasoning modes (2 new - Spoon / Einstein), and 5 instruct modes (2 new - Spoon / Einstein, all use ZERO REASONING TOKENS) all switchable on the fly via API, direct and “in chat” (yes - model ctrl at the chat/message level). Model name has “plusIQ” in the name. A 12+12 (12 reasoning and 12 instruct) model with interactive optimization/help system will be releasing shortly too. BF16/16-bit MTP GGUF also avail. (there is also a extra robust “tools” version too - Q6 and Q8.)

Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-NEO-IMATRIX-MAX-MTP-GGUF

Extreme intelligence in a small package. Jaw dropping performance. Superior instruction following. A multi-stage and multi-model fine tune and multi-stage merge on local hardware by myself and Nightmedia. Several of my 9B Qwen 3.5 fine tunes were used to make this model. The goal of this model was to raise the general intelligence of the model AND instruction following. Getting these right raises the performance of the model from top to bottom. Keep in mind, even thou this model EXCEEDS 7 critical benchmarks of Qwen 3.5 27B model, that does not mean it will beat the 27B in all tasks. This is also a HERETIC model, trained post “Heretic’ing” -> this model does what you want, no questions asked. Also: Thinking/reasoning block has been compacted, and in many cases it is a lot stronger. Fully uncensored. Benchmarks below. Example generations at the bottom of the page.

PS: If you are looking for a 27B model that BEATS Qwen 3.6 27B, Qwen 3.5 27B, and Qwen 35B-A3B, IN ALL critical benchmarks this is the one: https://huggingface.co/DavidAU/Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-NEO-MAX-MTP-GGUF

BENCHMARKS: (Bar Graph below.)

arc/c arc/e boolq hswag obkqa piqa wino
Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic [instruct] bf16 0.649,0.832,0.895,0.713,0.482,0.783,0.699
mxfp8 0.647,0.836,0.895,0.706,0.460,0.784,0.695
mxfp4 0.640,0.824,0.886,0.703,0.468,0.780,0.691
Qwen3.5-9B-Instruct: [base, non heretic] mxfp8 0.571,0.719,0.895,0.683,0.426,0.770,0.671
---
Qwen3.8-27B: [base, non heretic] mxfp8 0.591,0.782,0.896,0.746,0.448,0.801,0.711
mxfp4 0.581,0.771,0.889,0.738,0.442,0.798,0.713
Qwen3.6-27B-Instruct: [base, non heretic] mxfp8 0.647,0.803,0.910,0.773,0.450,0.806,0.742
Qwen3.5-27B-Instruct: [base, non heretic] mxfp8 0.557,0.711,0.868,0.533,0.452,0.706,0.695
Qwen3.6-35B-A3B-Instruct [base, non heretic] mxfp8 0.581,0.757,0.892,0.751,0.428,0.803,0.688

Graph note: First three bars are Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic.

“Bench” NOTES:

  • Models are tested in “Instruct” mode because this generally works better with the testing harness.
  • Testing via “thinking” mode also shows the metrics (and changes) but not the true extent.
  • In actual fact when the model IS in thinking mode, it will exceed INSTRUCT benchmark scores in most cases.

Regular and MTP GGUFS: Buy Me a Coffee (https://www.buymeacoffee.com/davidbelton)

All quants (regular and MTP) are NEO IMATRIX, which improve accuracy of the quants by an additional 2-4% over normal GGUFs as well as long context performance. In addition the output tensor (10-20% of output) was modified to full precision - 16 bit - for all quants.

“MTP” GGUFS (multi-token prediction):

  • “MTP” GGUFS will have “MTP” in the name as a suffix.
  • I have also set the MTP tensors to Q8_0 precision for all quants.
  • To get better performance keep temp 1 or less (higher temps degrade MTP performance).
  • Likewise with rep pen ; keep at 1 (off). If you raise it performance will suffer.
  • If you see “token acceptance” rates BELOW 50% (predict 2 tokens) switch to normal quants.

SPEED:

  • On Q4_K_S (4bit) quant, regular GGUFs are about 130 t/s, whereas MTP GGUFs (acceptance at 60%, 2 tokens) can exceed 185 T/S. (5090, Windows 11, testing in LMStudio)
  • Speeds will vary depending on GPU(s), AI app, O/S (Linux/Mac will generally be faster) and hardware.
  • “MTP” quants speeds will vary ; for creative/complex and/or temps over 1 use regular GGUFs for better performance. I suggest you download at least one of each - regular and MTP gguf(s) - and test them for your use case(s). If you get “token acceptance” (predict 2 tokens) with MTP quant(s) BELOW 50% (this means regular quants will run faster), then regular GGUF(s) will actually perform better - ie faster. MTP quant(s) can in some cases run faster as the token window fills up and/or in multi turn chats. Note there is NO other diffence between the quants type besides speed: both will do the same job.

Model:

  • 256k context
  • Gguf quants run in all standard AI apps.
  • Vision is activated, but you need to download separate “mmproj” file (ONE) to use it.

VISION:

  • Vision (images) tested.
  • You need an “mmproj” (just one) of these downloaded too, and placed in the same folder as the GGUF for images.

Qwen Model Settings (suggested):

  • Thinking mode for 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 for precise coding tasks (e.g. WebDev): temperature=0.6, top_p=0.95, top_k=20, min_p=0.0, presence_penalty=0.0, repetition_penalty=1.0
  • Instruct (or non-thinking) mode: temperature=0.7, top_p=0.80, top_k=20, min_p=0.0, presence_penalty=1.5, repetition_penalty=1.0
  • Context window min from 8k to 16k.

IMPORTANT - 5 Reasoning modes and 5 instruct modes:

GGUFS:

  • Model name has “plusIQ” in the name. (there is also a extra robust “tools” version too with “plusIQ-TOOLS” in the name.)
  • Avail in Q6 MTP, Q8 MTP and now: BF16-MTP [16 bit, full precision]

The good news is this: All the defaults are still the same for this model, that is “reasoning” is set at “xhigh” and if you activate “instruct mode” it will set automatically at “medium”. This was done to ensure “drop in” of this model into your workflow would work without issues/adjustments.

The GREAT NEWS is this: You are no longer limited to these defaults, and the both new reasoning modes and all modes of instruct are also unlocked too. Previously if you used “instruct” mode (thinking off) you were limited to only “medium” power with the model. We modified it so you now have “xhigh”, “medium” and “low” available too (as well as 2 new modes - more on that in a minute) We also modified the model so you can easily access and change between multiple reasoning and instruct modes at ANY TIME.

Here are the reasoning modes:

  • (There is no) “spoon” -> ULTRA xhigh, research mode // hyper detailed; this will automatically use more reasoning tokens (reasoning mode).
  • “einstein” -> a “high” mode that spawns up to 20 virtual agents to solve tasks; this will automatically use more reasoning tokens (reasoning mode).
  • “xhigh”, “medium” and “low” -> Standard Qwen 3.8 reasoning modes.

All these modes are also available via “instruct mode” too.

SPECIAL NOTE: “medium” is the default (both reasoning and instruct) for the model - IE no enhancement.

How strong are the enhancements? See 22k output here in this PDF, using Q6_K in “spoon” mode: Spoon Reasoning example PDF (https://huggingface.co/DavidAU/spoon-example1.pdf)

Next, we added coding to all “in message switching” (RIGHT IN CHAT):

{REASON:xxx} => Where “xxx” is “spoon”, “einstein”, “xhigh”, “medium” and “low”.

For instruct mode, just add an “i”: {REASON:xxx} => Where “xxx” is “ispoon”, “ieinstein”, “ixhigh”, “imedium” and “ilow”.

THE LAST “reason” / “instruct” mode will be the one used until you switch it again IN THE CURRENT CHAT.

EXAMPLES:

  • {REASON:einstein} tell me a story.
  • {REASON:ispoon} tell me a story.

Note the “{REASONxxx}” can be anywhere in the prompt, and will PERSIST until you change it again in the CHAT WINDOW/CURRENT CHAT. If you open a new chat window (depending on your AI APP) the DEFAULT reasoning mode at the default setting will take over unless you use the “{REASONxxx}” in the new prompt(s) at least ONCE in the new chat window/new chat session.

Also the systems automatically remove it from the “message stream” so the generation is “pure”.

For API this can be set manually - “reasoning” (you can use KW args to set this):

reasoning_effort = 'xhigh'
enable_thinking = 'true'

For API this can be set manually - “instruct” (you can use KW args to set this):

reasoning_effort = 'ixhigh'
enable_thinking = 'false'

ADVANCED: You can modify the jinja template to CHANGE the defaults by using the API code noted above. Place the defaults you want at the TOP of the jinja template.


DE-CENSORING:

https://huggingface.co/DavidAU/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-NEO-IMATRIX-MAX-MTP-GGUF#performance

Performance

KLD of less than 1 is excellent, zero is perfect.

MetricThis modelOriginal model (https://huggingface.co/Qwen/Qwen3.5-9B)
KL divergence0.07930*(by definition)***
Refusals6/100100/100

NOTES:

  • Suggest min q4ks (non-imatrix) or IQ3S (imatrix).
  • Tested with rep pen of 1 (off).
  • Context: 256k (default).

IMPORTANT:

  • Other versions in testing.
  • Information from Qwen’s repo below.
  • Video portions of the model were NOT TESTED.

The SUPER Qwen Universe - 40B, 27B and 9B ; meet the performance trendsetters:


NEW!: TURBO BRILLIANCE 50x: Reduce tokens, increase quality and IQ: The new tech from DavidAU to change and improve all models see the collection (3B to 120B; and growing fast): https://huggingface.co/collections/DavidAU/turbo-brilliance-ai-constructs-50x-reasoning-instruct

TURBO BRILLIANCE 50x: Reduce tokens, increase quality and IQ: The new tech from DavidAU to change and improve all models see the collection (3B to 120B; and growing fast): https://huggingface.co/collections/DavidAU/turbo-brilliance-ai-constructs-50x-reasoning-instruct

Qwen3.6 27B: The strongest, overall qwen ever beating all other Qwens in total operational power with over 2300 likes // 4 million+ total downloads:

  • https://huggingface.co/DavidAU/Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-NEO-MAX-MTP-GGUF

Qwen3.8 27B: The highest scoring Qwen in brute, raw intelligence, using Qwen 3.8’s 3 new reasoning modes, plus token reduction (1/2 to 1/10) enhancements:

  • https://huggingface.co/DavidAU/Qwen3.8-27B-TURBO-Fable-Cold-Fusion-735-882-Heretic-Uncensored-NEO-CODER-MAX-MTP-GGUF

Qwen3.8 27B: Super smart and 1/2 to 1/20 the reasoning tokens AND 5 reasoning/5 instruct modes switchable on the fly (even in chat):

  • https://huggingface.co/DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-GGUF

Qwen3.8 27B: 99% power of BF 16 at 4 and 8 bit. Power, Control and NO DE censoring for ultimate performance also with reasoning token reductions:

  • https://huggingface.co/DavidAU/Qwen3.8-27B-Cold-Fusion-GAIN-V1.1-NM-DAU-NEO-MAX-MTP-GGUF

Qwen3.6 40B: The 40B Monster, specializing in creative and research with 730+ likes and over 2 million downloads:

  • https://huggingface.co/DavidAU/Qwen3.6-40B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking-NEO-CODE-Di-IMatrix-MAX-GGUF

Qwen3.5 9B: At just 9B parameters it beats most untuned 27B models in both intelligence (640 ARC-C) and performance, plus features 5 reasoning and 5 instruct modes (Qwen 3.8) too:

  • https://huggingface.co/DavidAU/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-NEO-IMATRIX-MAX-MTP-GGUF

Using an “uncensored” (refusals removed) model VS trained “uncensored” model

Usually when you a tell a model to generate horror, swear or x-rated content this is all you have to do to get said content type. In the case of this model, it will not refuse your request, however it needs to be “pushed” a bit / directed a bit more in SOME CASES. Although this model will generated x-rated content too, likewise you need to tell it to use “slang” (and include the terms you want) to get it generate the content correctly as the “expected” content level too. Without these added directive(s), the content can be “bland” by comparison to an “uncensored model” or model trained on uncensored content. Roughly, the model tries to generate the content but the “default” setting(s) are so “tame” it needs a push to generate at expected graphic, cursing or explicit levels. Even with minimal direction (ie, use these words to swear: x,y,z), this will be enough to push the model to generate the requested content in the ahh… expected format.


Settings: CHAT / ROLEPLAY and/or SMOOTHER operation of this model:

In “KoboldCpp” or “oobabooga/text-generation-webui” or “Silly Tavern” ; Set the “Smoothing_factor” to 1.5 :

  • in KoboldCpp -> Settings->Samplers->Advanced-> “Smooth_F”
  • in text-generation-webui -> parameters -> lower right.
  • In Silly Tavern this is called: “Smoothing”

NOTE: For “text-generation-webui” -> if using GGUFs you need to use “llama_HF” (which involves downloading some config files from the SOURCE version of this model)

Source versions (and config files) of my models are here: https://huggingface.co/collections/DavidAU/d-au-source-files-for-gguf-exl2-awq-gptq-hqq-etc-etc-66b55cb8ba25f914cbf210be

OTHER OPTIONS:

  • Increase rep pen to 1.1 to 1.15 (you don’t need to do this if you use “smoothing_factor”)
  • If the interface/program you are using to run AI MODELS supports “Quadratic Sampling” (“smoothing”) just make the adjustment as noted.

Highest Quality Settings / Optimal Operation Guide / Parameters and Samplers

This a “Class 1” model: For all settings used for this model (including specifics for its “class”), including example generation(s) and for advanced settings guide (which many times addresses any model issue(s)), including methods to improve model performance for all use case(s) as well as chat, roleplay and other use case(s) please see: [https://huggingface.co/DavidAU/Maximizing-Model-Performance-All-Quants-Types-And-Full-Precision-by-Samplers_Parameters (https://huggingface.co/DavidAU/Maximizing-Model-Performance-All-Quants-Types-And-Full-Precision-by-Samplers_Parameters)]

You can see all parameters used for generation, in addition to advanced parameters and samplers to get the most out of this model here: [https://huggingface.co/DavidAU/Maximizing-Model-Performance-All-Quants-Types-And-Full-Precision-by-Samplers_Parameters (https://huggingface.co/DavidAU/Maximizing-Model-Performance-All-Quants-Types-And-Full-Precision-by-Samplers_Parameters)]

Similar Articles

@NFT_Chen: Mind-blowing! Qwen3.8-27B Hardcore Layer Expansion Into 40B! Just 21.8GB at 4bit for Local Fly-off! +50% Parameters, Layers Jumped From 64 to 96, Plus Claude Opus-Boosted Reasoning, Heretic Uncensored Tuning + Turbo Brilliance 50x AGI...

X AI KOLs Timeline

DavidAU released the Qwen3.8-40B Ultimate Heretic Uncensored model: it expands Qwen3.8-27B to 40B (from 64 layers to 96 layers), combines a Claude Opus reasoning dataset with Turbo Brilliance 50x AGI tuning, runs locally on an RTX 3090 at just 21.8GB in 4bit quantization, and supports Apache 2.0 open source as well as multimodal capabilities.

@KtAIFeed: Straight to the point, no fluff. The recently popular Qwen 3.6 (35B/43B) latest open-source 'uncensored' model on Hugging Face (over a million downloads per month) can run locally with just 6GB VRAM on a single GPU. It completely breaks the original model's moral preaching and safety restrictions—no censorship, it will answer whatever you ask...

X AI KOLs Timeline

Introduces the Qwen 3.6 (35B/43B) open-source uncensored model, removing official moral and safety restrictions. Requires only 6GB VRAM for local operation. Over a million downloads.

@sanbuphy: K2.6 successfully downloaded and deployed the Qwen3.5-0.8B model locally on a Mac, using the niche Zig language to implement and optimize inference, demonstrating the new model’s generalization ability. After 4,000+ tool calls and 12+ hours of continuous operation, K2.6 iterated 14 times…

X AI KOLs Timeline

K2.6 successfully downloaded and deployed the Qwen3.5-0.8B model locally on a Mac, using the niche Zig language to implement and optimize inference, demonstrating the new model’s generalization ability. After 4,000+ tool calls and 12+ hours of continuous operation, K2.6 iterated 14 times, boosting throughput from ~15 tokens/s to ~193 tokens/s, ultimately achieving 20% faster inference than LM Studio.