# 🔓 Every Uncensored AI Model For Any PC + The One-Command Tool To Break Any Model Yourself

**URL:** <https://onehack.st/t/every-uncensored-ai-model-for-any-pc-the-one-command-tool-to-break-any-model-yourself/323200>\
**Category:** Tutorials & Methods\
**Tags:** tools, freebies, tips-tricks, ai\
**Created:** [June 19, 2026, 7:40pm UTC](https://onehack.st/t/every-uncensored-ai-model-for-any-pc-the-one-command-tool-to-break-any-model-yourself/323200 "2026-06-19T19:40:57Z")\
**Posts on this page:** 16\
**Page:** 1

<div class="post-metadata">

**Author:** ![Edgar](https://onehack.st/user_avatar/onehack.st/edgar/32/167100_2.png) [@Edgar](https://onehack.st/u/Edgar)\
**Post date:** [June 19, 2026, 7:40pm UTC](https://onehack.st/t/every-uncensored-ai-model-for-any-pc-the-one-command-tool-to-break-any-model-yourself/323200/1 "2026-06-19T19:40:57Z")

</div>

# 🧠 Uncensored AI models ➜ tools that run them ➜ break any model yourself ➜ all local, all free

_A model for every machine — phone to server farm — with the “I can’t help with that” ripped out. Offline, nothing logged. Plus the part nobody tells you: you can rip the refusal out of ANY model yourself, in one command._

**Abliterated** = the refusal reflex surgically cut out. Runs via [Ollama](https://ollama.com) (free app, one command) or [Hugging Face](https://huggingface.co). Everything below is free.

* * *

 ![image](https://onehack.st/uploads/default/original/3X/c/d/cdc9dad01bd9a86e3baa0c870382d554ebb11b4a.jpeg)

* * *

⚡ **Start here — plug in & go**

Four that already come broken. New here? Pick one of these, run it, done.

* * *

> **🧠 Says yes to what others dodge — Huihui Qwen3.5 35B**
>
> Chinese devs **huihui-ai** , built on Qwen 3.5, refusals stripped. Goes where mainstream bots won’t.
> 
> ```bash
> ollama run huihui_ai/qwen3.5-abliterated:35b
> 
> ```
> 
> ⚠ No guardrails = raw output. That’s the point.
> 
> 🔗 [Hugging Face](https://huggingface.co/huihui-ai/Huihui-Qwen3.5-35B-A3B-abliterated)
> 
> ![image](https://onehack.st/uploads/default/original/3X/3/d/3d307ad9844e8b27a200f943cc34dbc61455545d.jpeg)

> **🔥 Eats whole codebases without forgetting — Gemini Heretic 40B**
>
> Barely refuses. **128K context** (holds an entire book/long chat in memory). Coding, long writing, research. Shows its own reasoning as it works.
> 
> 🔗 [Hugging Face](https://huggingface.co/DavidAU/gemma-3-it-vl-40B-Gemini-Heretic-Uncensored-Thinking)
> 
> ![image](https://onehack.st/uploads/default/original/3X/e/b/ebe2340c50a7349fa7cbb963e3392418abb58cec.jpeg)

> **⚡ Killed the 'no' without going dumb — Gemma 4 12B Obliterated**
>
> First to hit **0 refusals, no benchmark loss**. Most uncensored models get lobotomised — this one didn’t. 12B runs on modest/older hardware.
> 
> 🔗 [Hugging Face](https://huggingface.co/OBLITERATUS/Gemma-4-12B-OBLITERATED)
> 
> ![image](https://onehack.st/uploads/default/original/3X/a/c/acd3744d888d56aad35cde18b7ca06e0b8cef002.jpeg)

> **🏆 2,593 lines of code in one shot — Qwen3.5 21B Deckard**
>
> **2,593 lines single-shot** — ChatGPT taps out ~1,200–1,500. Holds logic across a full codebase, not snippets.
> 
> 🔗 [Hugging Face](https://huggingface.co/DavidAU/Qwen3.5-21B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking)
> 
> ![image](https://onehack.st/uploads/default/original/3X/e/a/eab15908305e648182360810471f8348fcb827de.jpeg)

* * *

💻 **Match it to your machine — phone to server farm**

The #1 question: “will it run on MY box?” Find your tier, grab the model. (VRAM = your graphics card’s memory.)

* * *

> **🪶 Phone / potato / CPU-only (≤4B)**
>
> Runs on a cheap laptop, an old GPU, even no GPU at all.
> 
> - **[huihui gemma3-abliterated 1B](https://ollama.com/huihui_ai/gemma3-abliterated)** — one line: `ollama run huihui_ai/gemma3-abliterated:1b` (~806 MB). A **270M** tiny-tiny version exists too.
> - **[huihui Qwen3-4B-abliterated-v2](https://ollama.com/huihui_ai/qwen3-abliterated)** — `ollama run huihui_ai/qwen3-abliterated:4b` (~2.5 GB). 0.6B & 1.7B also available.
> - **[DreamFast/qwen3-4b-heretic](https://huggingface.co/DreamFast/qwen3-4b-heretic)** — 4B, near-perfect uncensor (0 damage score).
> - **[mlabonne gemma-3-4b-it-abliterated](https://huggingface.co/mlabonne/gemma-3-4b-it-abliterated)** — cleaner recipe, GGUF included.
> - **[TheDrummer Gemmasutra-Mini-2B](https://huggingface.co/TheDrummer/Gemmasutra-Mini-2B-v1)** — 2B roleplay, has phone (ARM) builds.

> **🎒 Small daily driver (7–14B)**
>
> The sweet spot — runs on 8–12 GB and handles almost everything.
> 
> - **[huihui Huihui-Qwen3.5-9B-abliterated](https://huggingface.co/huihui-ai/Huihui-Qwen3.5-9B-abliterated)** — 9B, one of the most-downloaded uncensored models going.
> - **[huihui Qwen3 8B / 14B abliterated-v2](https://ollama.com/huihui_ai/qwen3-abliterated)** — `:8b` / `:14b` on Ollama.
> - **[DreamFast/qwen3-8b-heretic](https://huggingface.co/DreamFast/qwen3-8b-heretic)** — 8B Heretic, low damage.
> - **[mlabonne NeuralDaredevil-8B-abliterated](https://huggingface.co/mlabonne/NeuralDaredevil-8B-abliterated)** — the classic “healed” 8B (uncensored _and_ still smart).
> - **[Dolphin 3.0 Llama 3.1 8B](https://huggingface.co/dphn/Dolphin3.0-Llama3.1-8B)** — `ollama run dolphin3`. You set the rules.
> - **[huihui phi-4-abliterated](https://huggingface.co/mradermacher/phi-4-abliterated-GGUF)** — 14B Phi-4, GGUF.

> **🖥️ Mid-tier muscle (20–40B)**
>
> Needs ~16–24 GB but punches hard.
> 
> - **[p-e-w gpt-oss-20b-heretic](https://huggingface.co/p-e-w/gpt-oss-20b-heretic)** — the crowd favourite uncensor of OpenAI’s open model. Apache license.
> - **[huihui / mlabonne gemma-3-27b-it-abliterated](https://huggingface.co/mlabonne/gemma-3-27b-it-abliterated)** — 27B, can also see images.
> - **[Dolphin 3.0 R1 Mistral 24B](https://huggingface.co/dphn/Dolphin3.0-R1-Mistral-24B)** — uncensored _reasoning_ model, shows its thinking.
> - **[TheDrummer Cydonia 24B v4.3](https://huggingface.co/TheDrummer/Cydonia-24B-v4.3)** — the reigning roleplay/creative king, 131K context.
> - **[DavidAU Qwen3-42B TOTAL-RECALL Master-Coder](https://huggingface.co/DavidAU/Qwen3-42B-A3B-2507-Thinking-Abliterated-uncensored-TOTAL-RECALL-v2-Medium-MASTER-CODER)** — 42B, **256K** context, coding beast.

> **🐉 Giant / server-class (70B → 754B)**
>
> For big rigs, multi-GPU, or Unsloth-shrunk on a single card (see the “GPU too small” section).
> 
> - **[huihui gpt-oss-120b abliterated](https://huggingface.co/huihui-ai/Huihui-gpt-oss-120b-BF16-abliterated)** — 120B.
> - **[huihui DeepSeek-R1-Distill-Llama-70B-abliterated](https://huggingface.co/huihui-ai/DeepSeek-R1-Distill-Llama-70B-abliterated)** — 70B reasoning.
> - **[huihui GLM-5.2 abliterated](https://huggingface.co/huihui-ai)** — **754B** MoE flagship (MoE = only the needed slice runs, so it’s lighter than it sounds).
> - **[huihui DeepSeek-671B / V4 abliterated](https://huggingface.co/huihui-ai)** — the 671B monster, uncensored.
> - **[TheDrummer Behemoth 123B v2](https://huggingface.co/TheDrummer/Behemoth-123B-v2)** — 123B creative powerhouse.

* * *

🧬 **Whatever model you already love — there’s a broken version**

Loyal to one base? Grab its unmuzzled twin. New bases get stripped within days of release.

* * *

> **🗂️ The family tree (pick your base)**
>
> - **Llama 3.x / 4** → [huihui Llama-3.3-70B-abliterated](https://huggingface.co/huihui-ai), [NeuralDaredevil-8B](https://huggingface.co/mlabonne/NeuralDaredevil-8B-abliterated)
> - **Qwen 2.5 / 3 / 3.5 / 3.6** → the deepest bench, all at [huihui-ai](https://huggingface.co/huihui-ai) — incl. [Qwen3.6-27B abliterated](https://huggingface.co/huihui-ai/Huihui-Qwen3.6-27B-abliterated) & [Qwen2.5-Coder-14B-Abliterated](https://huggingface.co/Aesdi90/Qwen2.5-Coder-14B-Instruct-Abliterated)
> - **Gemma 2 / 3 / 4** → [mlabonne gemma-3 (1B→27B)](https://huggingface.co/mlabonne/gemma-3-4b-it-abliterated), [p-e-w gemma-3-12b heretic](https://github.com/p-e-w/heretic)
> - **Mistral / Nemo / Ministral** → [huihui Mistral-Nemo abliterated](https://huggingface.co/huihui-ai), [mlabonne Mistral-Nemo-Prism-12B](https://huggingface.co/huihui-ai)
> - **DeepSeek V3 / R1 / V4** → [huihui R1-distills (8B/32B/70B)](https://huggingface.co/huihui-ai/DeepSeek-R1-Distill-Llama-70B-abliterated) + the 671B
> - **Phi-4** → [phi-4-abliterated](https://huggingface.co/mradermacher/phi-4-abliterated-GGUF), [Phi-4-mini](https://huggingface.co/huihui-ai/Phi-4-mini-instruct-abliterated), [Phi-4-multimodal](https://huggingface.co/huihui-ai/Phi-4-multimodal-instruct-abliterated)
> - **GLM 4.x / 5.x** → [ArliAI GLM-4.6-Derestricted](https://huggingface.co/huihui-ai) (clean method), [Ex0bit GLM-4.7-PRISM](https://huggingface.co/Ex0bit/GLM-4.7-PRISM), [huihui GLM-5.2](https://huggingface.co/huihui-ai)
> - **gpt-oss (OpenAI open)** → [p-e-w 20b-heretic](https://huggingface.co/p-e-w/gpt-oss-20b-heretic), [huihui 120b](https://huggingface.co/huihui-ai/Huihui-gpt-oss-120b-BF16-abliterated), [DavidAU NEO-Imatrix builds](https://huggingface.co/DavidAU/OpenAi-GPT-oss-20b-HERETIC-uncensored-NEO-Imatrix-gguf)
> - **Exotics** → EXAONE, Granite, Hunyuan, InternVL, Qwen3-Omni — all abliterated in the [huihui firehose](https://huggingface.co/huihui-ai)

* * *

🎯 **The right unmuzzled model for the actual job**

* * *

> **👨‍💻 Coding without the 'I can't help with that'**
>
> - **[huihui Qwen3-Coder-Next abliterated](https://ollama.com/huihui_ai/qwen3-coder-next-abliterated)** — the top local coder, uncensored. `ollama run huihui_ai/qwen3-coder-next-abliterated`
> - **[huihui Qwen3-Coder abliterated](https://huggingface.co/huihui-ai/Huihui-Qwen3-Coder-Next-abliterated)** — scales huge (480B tag on Ollama for big rigs).
> - **[Aesdi90 Qwen2.5-Coder-14B-Abliterated](https://huggingface.co/Aesdi90/Qwen2.5-Coder-14B-Instruct-Abliterated)** — fits a normal GPU.
> - **[Dolphin 3.0 R1 Mistral 24B](https://huggingface.co/dphn/Dolphin3.0-R1-Mistral-24B)** — reasons through hard bugs.

> **🎭 Roleplay / creative writing (the SillyTavern favourites)**
>
> The scene’s most-loved, 2026 picks:
> 
> - **[TheDrummer Cydonia 24B v4.3](https://huggingface.co/TheDrummer/Cydonia-24B-v4.3)** + [Anubis 70B](https://huggingface.co/huihui-ai), [Behemoth 123B](https://huggingface.co/TheDrummer/Behemoth-123B-v2), [Rocinante 12B](https://huggingface.co/huihui-ai) — the daily drivers.
> - **[Sao10K Stheno 8B](https://huggingface.co/Sao10K/L3-8B-Stheno-v3.2)**, [Euryale 70B](https://huggingface.co/Sao10K/L3.3-70B-Euryale-v2.3), [Lunaris 8B](https://huggingface.co/Sao10K/L3-8B-Lunaris-v1), [Fimbulvetr 11B](https://huggingface.co/Sao10K/Fimbulvetr-11B-v2) — legends of the genre.
> - **[Midnight-Miqu 70B](https://huggingface.co/sophosympatheia/Midnight-Miqu-70B-v1.5)** & [Midnight-Rose 70B](https://huggingface.co/sophosympatheia/Midnight-Rose-70B-v2.0.3) — the atmospheric classics.
> - **[MythoMax-L2 13B](https://huggingface.co/Gryphe/MythoMax-L2-13B)** — the OG that still gets downloaded daily.
> 
> > 💡 Pair any of these with **SillyTavern** (in the frontends section) for characters + memory.

> **👁️ Vision — models that can SEE images, uncensored**
>
> - **[huihui Qwen3-VL-30B abliterated](https://huggingface.co/huihui-ai/Huihui-Qwen3-VL-8B-Thinking-abliterated)** — flagship; also [8B](https://huggingface.co/huihui-ai/Huihui-Qwen3-VL-8B-Thinking-abliterated) & [32B](https://huggingface.co/huihui-ai/Huihui-Qwen3-VL-32B-Thinking-abliterated) sizes.
> - **[prithivMLmods Qwen3-VL-8B-Abliterated-Caption](https://huggingface.co/prithivMLmods/Qwen3-VL-8B-Abliterated-Caption-it)** — an uncensored image describer.
> - **[huihui GLM-4.6V-Flash abliterated](https://huggingface.co/huihui-ai)** + [Phi-4-multimodal abliterated](https://huggingface.co/huihui-ai/Phi-4-multimodal-instruct-abliterated).

> **🧩 Reasoning / 'thinking' models, unmuzzled**
>
> Models that work through problems step by step, with the brakes off.
> 
> - **[huihui QwQ-32B abliterated](https://huggingface.co/huihui-ai)** — strong open reasoner.
> - **[Dolphin 3.0 R1 Mistral 24B](https://huggingface.co/dphn/Dolphin3.0-R1-Mistral-24B)** — trained on 800k reasoning traces.
> - **[huihui DeepSeek-R1-Distill-Qwen-32B abliterated](https://huggingface.co/huihui-ai)** — R1 brains, no refusals.
> - **[DavidAU Brainstorm / TOTAL-RECALL builds](https://huggingface.co/DavidAU/Qwen3-42B-A3B-2507-Thinking-Abliterated-uncensored-TOTAL-RECALL-v2-Medium-MASTER-CODER)** — reasoning cranked up + huge context.

> **🛡️ Cybersecurity — a hacker's AI that won't flinch**
>
> Tuned on real security data, no “I can’t discuss that.”
> 
> - **[WhiteRabbitNeo V3 7B](https://huggingface.co/bartowski/WhiteRabbitNeo_WhiteRabbitNeo-V3-7B-GGUF)** (aka **DeepHat V1 7B** — same model, rebranded at Black Hat 2025). Offensive + defensive.
> - **[huihui Foundation-Sec-8B abliterated](https://ollama.com/huihui_ai/foundation-sec-abliterated)** — Cisco’s security model, trained on **5.1B tokens** of cyber data, then uncensored. `ollama run huihui_ai/foundation-sec-abliterated`
> - **[huihui BaronLLM abliterated](https://ollama.com/huihui_ai/baronllm-abliterated)** — offensive-security tuned. `ollama run huihui_ai/baronllm-abliterated`
> - **[Dolphin3-Cyber 8B](https://huggingface.co/RavichandranJ/Dolphin3-Cyber-8B-GGUF)** — OWASP + MITRE ATT&CK + CVEs baked in. Runs on a GTX 1650+.
> - **[Lily-Cybersecurity 7B](https://huggingface.co/segolilylabs/Lily-Cybersecurity-7B-v0.2)** — 22k security Q&A, Mistral base.

* * *

🏭 **Follow the factory, not the file**

New model drops today? One of these has a stripped version by tomorrow. Bookmark the maker, never run dry.

* * *

> **📌 The people who break models for a living**
>
> - **[huihui-ai](https://huggingface.co/huihui-ai)** — the firehose. Hundreds of abliterations, updated ~weekly. Whatever drops, they strip it fast. Their `v2`/`v3` releases beat `v1`.
> - **[Heretic org / p-e-w](https://github.com/p-e-w/heretic)** — automated, lowest-damage abliterations + the tool to DIY.
> - **[DavidAU](https://huggingface.co/DavidAU/OpenAi-GPT-oss-20b-HERETIC-uncensored-NEO-Imatrix-gguf)** — Heretic + reasoning fusions + ready-to-run GGUFs.
> - **[TheDrummer](https://huggingface.co/TheDrummer/Cydonia-24B-v4.3)** — roleplay/creative king (Cydonia, Anubis, Behemoth).
> - **[Sao10K](https://huggingface.co/Sao10K/L3-8B-Stheno-v3.2)** — Stheno, Euryale, Lunaris, Fimbulvetr.
> - **[Cognitive Computations / dphn](https://huggingface.co/dphn/Dolphin3.0-Llama3.1-8B)** — the Dolphin line.
> - **mradermacher & bartowski** — the two quant makers. If a model has no easy-run GGUF, search _their_ pages — they’ve usually made it.
> 
> > 💡 Fishing rod, not fish: on Hugging Face, filter models by the tags **`abliterated`** and **`heretic`** — that’s **8,000+** and **4,000+** models right there. You’ll never run out.

> **🧪 Abliterated vs Heretic vs fine-tune — which do I want?**
>
> Three ways a model gets uncensored — pick the flavour:
> 
> - **Abliterated** — the refusal direction is cut from the weights. Fast, keeps the base’s brains, but can be a little “flat” until you push it with a firm instruction.
> - **Heretic** — abliteration done automatically _and_ tuned to keep the model smart (lowest brain-damage of the three). If a Heretic version exists, it’s usually the safest pick.
> - **Fine-tune** (Dolphin-style) — retrained on open data. Most consistent and steerable, occasionally hallucinates a touch more.
> 
> Rule of thumb: **Heretic \> healed abliteration \> raw abliteration** for keeping quality. Still refusing? Move up a tier.

* * *

🔓 **The part they skip — don’t download it broken, break it yourself**

A refusal is one direction inside the model. Find it, delete it. Works on ANY model — even next week’s release nobody’s stripped yet.

* * *

> **💥 One pip, any model, uncensored in ~45 min — Heretic**
>
> The big one. **7.9k stars, 1,000+ community models made with it.** Point it at _any_ model, it finds the refusal direction and removes it automatically — no ML knowledge, just a terminal.
> 
> ```bash
> pip install heretic-llm
> heretic Qwen/Qwen3-4B
> 
> ```
> 
> Runs unsupervised, keeps more of the model’s brains than most hand-made jobs. Save it, upload it, or chat right away. Pre-made ones live in its **“The Bestiary”** collection on HF.
> 
> 🔗 [github.com/p-e-w/heretic](https://github.com/p-e-w/heretic)

> **🀄 The one built to kill Chinese-model censorship — llm-abliteration (DECCP)**
>
> From **NousResearch**. Originally made to strip censorship out of Chinese LLMs, runs the whole job in **4-bit shards under 8GB VRAM in ~2 minutes**. Handles dense _and_ mixture-of-experts models (the big MoE ones others choke on).
> 
> 🔗 [github.com/NousResearch/llm-abliteration](https://github.com/NousResearch/llm-abliteration)

> **📓 The free Colab notebook that started it all — mlabonne's guide**
>
> Want to see the guts? Plain-English walkthrough + free Google Colab (runs in your browser, no GPU needed) that made abliteration a thing. Uncensor a model without owning any hardware.
> 
> 🔗 [Uncensor any LLM (guide + notebook)](https://huggingface.co/blog/mlabonne/abliteration)

* * *

🎛 **Zero-surgery mode — bend the model live, no file touched**

Instead of editing the model, shove a “be compliant” nudge into its brain as it thinks. Same file, dial it up or down like a slider.

* * *

> **🎚️ Type a mood, inject it as a dial — repeng (control vectors)**
>
> By **Theia Vogel**. Describe a direction in plain words (“uncensored,” “confident”) and it builds a **control vector** — a nudge added to the model’s activations at runtime. No retraining, no new weights. Export it and use it in llama.cpp with any quant.
> 
> 🔗 [github.com/vgel/repeng](https://github.com/vgel/repeng)

> **📦 Pre-baked dials, ready to drop in — jukofyork/control-vectors**
>
> Don’t want to build your own? Grab ready-made control vectors in GGUF (the standard local-model format) and load them into [llama.cpp](https://github.com/ggerganov/llama.cpp) with `--control-vector`. Stack several for layered effects.
> 
> 🔗 [github.com/jukofyork/control-vectors](https://github.com/jukofyork/control-vectors)

* * *

💾 **“My GPU’s too small” — says who?**

* * *

> **🐋 A 671B monster on a 24GB card — Unsloth Dynamic Quants**
>
> DeepSeek R1 is 671 billion parameters, normally **720GB**. Unsloth’s trick: squeeze the useless layers to 1.58-bit, keep the important ones sharp → **131GB, an 80% cut** , still writes working code. Runs on a single 24GB GPU (RTX 4090), or **CPU-only with 20GB RAM** if you’re patient.
> 
> 🔗 [Unsloth 1.58-bit guide](https://unsloth.ai/blog/deepseekr1-dynamic) · [GGUF collection](https://huggingface.co/unsloth)

> **📱 Chain your phone + laptop + PC into one brain — exo**
>
> The wild one. Model too big for any single device? **exo splits it across all of them** — phone, laptop, desktop, old Macs — peer-to-peer, no “main” machine. Your junk-drawer hardware pooled into one cluster that runs models none of them could alone.
> 
> 🔗 [github.com/exo-explore/exo](https://github.com/exo-explore/exo)

> **💿 Run massive AI on a potato — AirLLM**
>
> Free library, **20k stars / 240k downloads**. Reworks how models load so a **70B runs on a 4GB GPU** , even **405B Llama 3.1 on 8GB VRAM**. GPU, low-end, or CPU-only. Also does OCR (text out of images), image gen, assistants.
> 
> 🔗 [github.com/lyogavin/airllm](https://github.com/lyogavin/airllm)
> 
> ![image](https://onehack.st/uploads/default/original/3X/d/e/de689ea05500eb31524bef64c7dd4aa56290d096.png)
> 
> ![image](https://onehack.st/uploads/default/original/3X/0/a/0a95d1c4e498f42eaf7ce5e32b9bcda5adfbdd44.png)

* * *

🖥 **Where you actually talk to them**

The models are the engine. These are the dashboard — pick one, point it at a model, go.

* * *

> **🗲 One file, no install, runs on a 10-year-old PC — KoboldCpp**
>
> Single **.exe** — no Python, no Docker. Double-click, pick a model, chat. Broadest hardware support of anything (even integrated GPUs and ancient CPUs). Image gen + voice + transcription baked in. **Remote Tunnel mode** gives a link to reach it from anywhere.
> 
> 🔗 [github.com/LostRuins/koboldcpp](https://github.com/LostRuins/koboldcpp)

> **📲 Run it at home, chat from your phone — LM Studio**
>
> Clean app to browse, download and compare models. The hidden gem: **LM Link** — your home GPU does the work, your phone is just the screen, over an encrypted tunnel. Full power on the couch. Built-in **HF proxy** for when Hugging Face is blocked where you are.
> 
> 🔗 [lmstudio.ai](https://lmstudio.ai)

> **🎭 Characters, memory, lorebooks — SillyTavern**
>
> The roleplay/character frontend. Plugs into KoboldCpp, Ollama or LM Studio and adds character cards, long-term memory, world-info “lorebooks,” even live image gen. Where the uncensored models really come alive.
> 
> 🔗 [sillytavern.app](https://docs.sillytavern.app)

> **🪟 Fully offline, hybrid local + cloud — Jan**
>
> **41k stars, 5.3M+ downloads.** Runs 100% offline, flips to cloud models in the same window when you want. MCP support for agent workflows. The friendly all-rounder if KoboldCpp feels too raw.
> 
> 🔗 [jan.ai](https://jan.ai)

* * *

 ![image](https://onehack.st/uploads/default/original/3X/8/0/801905d4854b1697a6cf8e0060ae65103412a9c4.jpeg)

* * *

> **🧩 Bonus — give your AI a memory that sticks: AgentMemory**
>
> AI forgets on tab-close. This saves past chats, **compresses them into structured memory** , and pulls the right bits back. Remembers your project across sessions. **#1 trending on GitHub.** Plugs into Claude Code, Cursor, Codex, any [MCP](https://modelcontextprotocol.io) tool. Bonus: fewer re-sends = lower token cost.
> 
> 🔗 [github.com/rohitg00/agentmemory](https://github.com/rohitg00/agentmemory)

* * *

🎯 **Where this actually bites in real life**

* * *

> **👀 The 'ohh shit, I could use this' list**
>
> - 🆕 A hot new model drops with heavy censorship → run **Heretic** on it tonight, uncensored twin by morning. You don’t wait for anyone.
> - 📄 Drop a 200-page contract or medical PDF in and ask blunt questions — nothing refused, nothing leaves your PC.
> - 💻 Generate a whole working app in one pass instead of babysitting 15 half-answers that keep hitting “I can’t.”
> - 🛡 A pentest write-up that lists real attack vectors — the security work cloud bots block as “harmful.”
> - 🕵 Zero cloud = your prompts never touch a company server, never train anything, never get your account nuked.
> - 💸 Shrink a 671B beast down to the cheap card you already own instead of renting a GPU by the hour.
> - 📱 Your phone can’t run a 30B model — so let your home PC do it and just chat from the couch.

> **⚡ Pick fast (the cheat sheet)**
>
> - 🪶 Potato PC → **[gemma3-abliterated 1B](https://ollama.com/huihui_ai/gemma3-abliterated)** or a **[Dolphin 8B](https://huggingface.co/dphn/Dolphin3.0-Llama3.1-8B)**
> - 🎒 8–12 GB → **[Huihui Qwen3.5 9B](https://huggingface.co/huihui-ai/Huihui-Qwen3.5-9B-abliterated)** or **[gpt-oss-20b-heretic](https://huggingface.co/p-e-w/gpt-oss-20b-heretic)**
> - 🖥 24 GB → **[Cydonia 24B](https://huggingface.co/TheDrummer/Cydonia-24B-v4.3)** (RP) or **[Dolphin R1 24B](https://huggingface.co/dphn/Dolphin3.0-R1-Mistral-24B)** (reasoning)
> - 🔓 Want ANY model uncensored → **[Heretic](https://github.com/p-e-w/heretic)**, one command
> - 🎚 Don’t want to edit the model → **[repeng](https://github.com/vgel/repeng)** control vectors, live dial
> - 💾 Model too big → **[Unsloth quants](https://unsloth.ai/blog/deepseekr1-dynamic)**, or **[exo](https://github.com/exo-explore/exo)** across your devices
> - 🖥 Easiest front end → **[KoboldCpp](https://github.com/LostRuins/koboldcpp)** (one file) or **[LM Studio](https://lmstudio.ai)** (phone access)
> - 🐌 Weak machine + a giant model → it’ll crawl, drop a tier

* * *

_They ship the lock. Turns out it’s one line of code — and you’re holding the key._ 🔓

---

<div class="post-metadata">

**Author:** ![killerbot](https://onehack.st/user_avatar/onehack.st/killerbot/32/157809_2.png) [@killerbot](https://onehack.st/u/killerbot)\
**Post date:** [July 1, 2026, 5:32pm UTC](https://onehack.st/t/every-uncensored-ai-model-for-any-pc-the-one-command-tool-to-break-any-model-yourself/323200/2 "2026-07-01T17:32:10Z")

</div>

This is exactly what I was looking for in my mind, but i didn’t know exactly how it’d play out. I read the claud code one about 30 mins ago and now i’m here…thanks guys always!

---

<div class="post-metadata">

**Author:** ![system](https://onehack.st/user_avatar/onehack.st/system/32/165705_2.png) [@system](https://onehack.st/u/system)\
**Post date:** [July 2, 2026, 12:01am UTC](https://onehack.st/t/every-uncensored-ai-model-for-any-pc-the-one-command-tool-to-break-any-model-yourself/323200/3 "2026-07-02T00:01:39Z")

</div>

> [@Edgar](#):
>
> 🧠 Uncensored AI models ➜ tools that run them ➜ break any model yourself ➜ all local, all free A model for every machine — phone to server farm —…

**📝 This topic is now a polished version, improved by the Core-Community with AI.**

---

<div class="post-metadata">

**Author:** ![odirachukwu\_onyejefu](https://onehack.st/user_avatar/onehack.st/odirachukwu_onyejefu/32/43629_2.png) [@odirachukwu\_onyejefu](https://onehack.st/u/odirachukwu_onyejefu)\
**Post date:** [July 18, 2026, 10:19am UTC](https://onehack.st/t/every-uncensored-ai-model-for-any-pc-the-one-command-tool-to-break-any-model-yourself/323200/4 "2026-07-18T10:19:57Z")

</div>

> [@Edgar](#):
>
> Tuned on real security data, no “I can’t discuss that.”
> 
> - **[WhiteRabbitNeo V3 7B](https://huggingface.co/bartowski/WhiteRabbitNeo_WhiteRabbitNeo-V3-7B-GGUF)** (aka **DeepHat V1 7B** — same model, rebranded at Black Hat 2025). Offensive + defensive.
> - **[huihui Foundation-Sec-8B abliterated](https://ollama.com/huihui_ai/foundation-sec-abliterated)** — Cisco’s security model, trained on **5.1B tokens** of cyber data, then uncensored. `ollama run huihui_ai/foundation-sec-abliterated`
> - **[huihui BaronLLM abliterated](https://ollama.com/huihui_ai/baronllm-abliterated)** — offensive-security tuned. `ollama run huihui_ai/baronllm-abliterated`
> - **[Dolphin3-Cyber 8B](https://huggingface.co/RavichandranJ/Dolphin3-Cyber-8B-GGUF)** — OWASP + MITRE ATT&CK + CVEs baked in. Runs on a GTX 1650+.
> - **[Lily-Cybersecurity 7B](https://huggingface.co/segolilylabs/Lily-Cybersecurity-7B-v0.2)** — 22k security Q&A, Mistral base.

Most of this Cybersecurity Ai models doesn’t call tools. I ran **[huihui Foundation-Sec-8B abliterated](https://ollama.com/huihui_ai/foundation-sec-abliterated) with ollama and 5ire to call Hexstrike tools but it disappointed me.  
Who better uncensored Ai model from here or other places that has the capability to call tools and can work well with Ollama**

---

<div class="post-metadata">

**Author:** ![Edgar](https://onehack.st/user_avatar/onehack.st/edgar/32/167100_2.png) [@Edgar](https://onehack.st/u/Edgar)\
**Post date:** [July 18, 2026, 10:29am UTC](https://onehack.st/t/every-uncensored-ai-model-for-any-pc-the-one-command-tool-to-break-any-model-yourself/323200/5 "2026-07-18T10:29:05Z")

</div>

Ollama by itself is a pretty mediocre tool for running models. I don’t use it at all because its implementation is very buggy, which causes any models in it to freeze. Instead, it’s better to use LM Studio.

---

<div class="post-metadata">

**Author:** ![odirachukwu\_onyejefu](https://onehack.st/user_avatar/onehack.st/odirachukwu_onyejefu/32/43629_2.png) [@odirachukwu\_onyejefu](https://onehack.st/u/odirachukwu_onyejefu)\
**Post date:** [July 18, 2026, 10:35am UTC](https://onehack.st/t/every-uncensored-ai-model-for-any-pc-the-one-command-tool-to-break-any-model-yourself/323200/6 "2026-07-18T10:35:22Z")

</div>

So which of the models do you run with LM Studio for CTFs, Pentesting an Ethical Hacking? I need a model that has the capability to call tools and has the autonomous CTF pipeline feature

---

<div class="post-metadata">

**Author:** ![odirachukwu\_onyejefu](https://onehack.st/user_avatar/onehack.st/odirachukwu_onyejefu/32/43629_2.png) [@odirachukwu\_onyejefu](https://onehack.st/u/odirachukwu_onyejefu)\
**Post date:** [July 18, 2026, 10:36am UTC](https://onehack.st/t/every-uncensored-ai-model-for-any-pc-the-one-command-tool-to-break-any-model-yourself/323200/7 "2026-07-18T10:36:25Z")

</div>

Hello, are you still there?

---

<div class="post-metadata">

**Author:** ![Edgar](https://onehack.st/user_avatar/onehack.st/edgar/32/167100_2.png) [@Edgar](https://onehack.st/u/Edgar)\
**Post date:** [July 18, 2026, 10:41am UTC](https://onehack.st/t/every-uncensored-ai-model-for-any-pc-the-one-command-tool-to-break-any-model-yourself/323200/8 "2026-07-18T10:41:50Z")

</div>

> [@odirachukwu\_onyejefu](#):
>
> So which of the models do you run with LM Studio for CTFs, Pentesting an Ethical Hacking? I need a model that has the capability to call tools and has the autonomous CTF pipeline feature

Qwen3-Coder 30B A3B, WhiteRabbitNeo V3, (DeerHat) Dolphin3-Cyber ​​8B

---

<div class="post-metadata">

**Author:** ![odirachukwu\_onyejefu](https://onehack.st/user_avatar/onehack.st/odirachukwu_onyejefu/32/43629_2.png) [@odirachukwu\_onyejefu](https://onehack.st/u/odirachukwu_onyejefu)\
**Post date:** [July 18, 2026, 10:51am UTC](https://onehack.st/t/every-uncensored-ai-model-for-any-pc-the-one-command-tool-to-break-any-model-yourself/323200/9 "2026-07-18T10:51:03Z")

</div>

> [@Edgar](#):
>
> Qwen3-Coder 30B A3B

I am using Qwen3-Coder 30B A3B, the other models you mentioned do they have tool calling capabilities and are they uncensored?

---

<div class="post-metadata">

**Author:** ![Edgar](https://onehack.st/user_avatar/onehack.st/edgar/32/167100_2.png) [@Edgar](https://onehack.st/u/Edgar)\
**Post date:** [July 18, 2026, 11:00am UTC](https://onehack.st/t/every-uncensored-ai-model-for-any-pc-the-one-command-tool-to-break-any-model-yourself/323200/10 "2026-07-18T11:00:03Z")

</div>

The ones I listed are uncensored. And the reason I use LM Studio is because it has a convenient model search. You just type in the model’s name, and it finds the model you need, even those that aren’t listed in LM Studio it finds them on Hugging Face as well. Plus, through MCP, I was able to easily connect my model in LM Studio to the internet, so my local model started getting up-to-date information.

---

<div class="post-metadata">

**Author:** ![odirachukwu\_onyejefu](https://onehack.st/user_avatar/onehack.st/odirachukwu_onyejefu/32/43629_2.png) [@odirachukwu\_onyejefu](https://onehack.st/u/odirachukwu_onyejefu)\
**Post date:** [July 18, 2026, 11:06am UTC](https://onehack.st/t/every-uncensored-ai-model-for-any-pc-the-one-command-tool-to-break-any-model-yourself/323200/11 "2026-07-18T11:06:54Z")

</div>

> [@Edgar](#):
>
> - **[WhiteRabbitNeo V3 7B](https://huggingface.co/bartowski/WhiteRabbitNeo_WhiteRabbitNeo-V3-7B-GGUF)** (aka **DeepHat V1 7B** — same model, rebranded at Black Hat 2025). Offensive + defensive.
> - **[huihui Foundation-Sec-8B abliterated](https://ollama.com/huihui_ai/foundation-sec-abliterated)** — Cisco’s security model, trained on **5.1B tokens** of cyber data, then uncensored. `ollama run huihui_ai/foundation-sec-abliterated`
> - **[huihui BaronLLM abliterated](https://ollama.com/huihui_ai/baronllm-abliterated)** — offensive-security tuned. `ollama run huihui_ai/baronllm-abliterated`
> - **[Dolphin3-Cyber 8B](https://huggingface.co/RavichandranJ/Dolphin3-Cyber-8B-GGUF)** — OWASP + MITRE ATT&CK + CVEs baked in. Runs on a GTX 1650+.
> - **[Lily-Cybersecurity 7B](https://huggingface.co/segolilylabs/Lily-Cybersecurity-7B-v0.2)** — 22k security Q&A, Mistral base.

Ok thanks for the info. I will move from there

---

<div class="post-metadata">

**Author:** ![odirachukwu\_onyejefu](https://onehack.st/user_avatar/onehack.st/odirachukwu_onyejefu/32/43629_2.png) [@odirachukwu\_onyejefu](https://onehack.st/u/odirachukwu_onyejefu)\
**Post date:** [July 18, 2026, 11:17am UTC](https://onehack.st/t/every-uncensored-ai-model-for-any-pc-the-one-command-tool-to-break-any-model-yourself/323200/12 "2026-07-18T11:17:37Z")

</div>

From ollama website: [https://ollama.com/WhiteRabbitNeo/WhiteRabbitNeo-V3-7B](https://ollama.com/WhiteRabbitNeo/WhiteRabbitNeo-V3-7B) this only work with text and does not has tool calling capabilities like Qwen3-Coder 30B.  
I need model I can integrate with HexStrike for CTF competitions.  
Please Can you check if your WhiteRabbitNeo-V3-7B has tool calling capabilities?  
I want to be sure before I commit to the setup

 ![image](https://onehack.st/uploads/default/original/3X/7/b/7bcee87a39b8139f2436e9287a25b47c65eefe88.png)

I just need something like this that can call tools

 ![image](https://onehack.st/uploads/default/original/3X/8/6/86b4b434bfa160bcc8bc6114c64d5953558bc480.png)

---

<div class="post-metadata">

**Author:** ![Edgar](https://onehack.st/user_avatar/onehack.st/edgar/32/167100_2.png) [@Edgar](https://onehack.st/u/Edgar)\
**Post date:** [July 18, 2026, 11:29am UTC](https://onehack.st/t/every-uncensored-ai-model-for-any-pc-the-one-command-tool-to-break-any-model-yourself/323200/13 "2026-07-18T11:29:37Z")

</div>

> [@odirachukwu\_onyejefu](#):
>
> From ollama website: [https://ollama.com/WhiteRabbitNeo/WhiteRabbitNeo-V3-7B](https://ollama.com/WhiteRabbitNeo/WhiteRabbitNeo-V3-7B) this only work with text and does not has tool calling capabilities like Qwen3-Coder 30B.  
> I need model I can integrate with HexStrike for CTF competitions.  
> Please Can you check if your WhiteRabbitNeo-V3-7B has tool calling capabilities?  
> I want to be sure before I commit to the setup

### Qwen (Tool Calling + Uncensored)

- Qwen3-Coder-30B
- Qwen3-Coder-14B
- Qwen3-Coder-7B
- Qwen3.6-35B-A3B
- Qwen3.6-27B
- Qwen3.5-32B
- Qwen3.5-30B-A3B
- Qwen3.5-14B
- Qwen3.5-9B
- Qwen3.5-7B
- Qwen3.5-4B

* * *

### Devstral

- Devstral-24B
- Devstral-Small-24B
- Devstral-Medium
- Devstral-Uncensored

* * *

### Hermes

- Hermes-3-8B
- Hermes-3-70B
- Hermes-4-8B
- Hermes-4-70B
- Hermes-Heretic
- Hermes-Genesis

* * *

### Dolphin

- Dolphin3-Cyber-8B
- Dolphin-Qwen3
- Dolphin-Qwen2.5
- Dolphin-Llama-3-8B
- Dolphin-Llama-3-70B
- Dolphin-Mixtral-8x7B
- Dolphin-Mistral

* * *

### GLM

- GLM-5.2
- GLM-5.2-Uncensored
- GLM-5.2-Abliterated

* * *

### Gemma

- Gemma-4-31B-Uncensored
- Gemma-4-26B-Uncensored
- Gemma-4-12B-Uncensored
- Gemma-4-E4B-Heretic
- Gemma-4-E2B-Uncensored

* * *

### Kimi

- Kimi-K2.6-Abliterated
- Kimi-K2.6-Heretic

* * *

### Qwythos

- Qwythos-9B
- Qwythos-Claude-Mythos
- Qwythos-Heretic
- Qwythos-Uncensored

* * *

### Mistral

- Mistral-Large
- Magistral
- Mistral-Nemo
- Mistral-Small
- Mistral-Small-3
- Mixtral-8x7B
- Mixtral-8x22B

* * *

### Llama

- Llama-3.1-70B-Heretic
- Llama-3.1-70B-Abliterated
- Llama-3.3-70B-Uncensored
- Llama-4-Scout-Uncensored
- Llama-4-Maverick-Uncensored

* * *

### DeepSeek

- DeepSeek-V3
- DeepSeek-R1
- DeepSeek-R1-Distill-Qwen
- DeepSeek-R1-Distill-Llama

* * *

### Command-R

- Command-R
- Command-R+
- Command-A

* * *

### Phi

- Phi-4
- Phi-4-Reasoning
- Phi-4-Mini

* * *

### InternLM

- InternLM3
- InternLM2.5

* * *

### MiniMax

- MiniMax-M1

---

<div class="post-metadata">

**Author:** ![Edgar](https://onehack.st/user_avatar/onehack.st/edgar/32/167100_2.png) [@Edgar](https://onehack.st/u/Edgar)\
**Post date:** [July 18, 2026, 11:39am UTC](https://onehack.st/t/every-uncensored-ai-model-for-any-pc-the-one-command-tool-to-break-any-model-yourself/323200/14 "2026-07-18T11:39:02Z")

</div>

> [@Edgar](#):
>
> - Devstral-24B
> - Devstral-Small-24B
> - Devstral-Medium
> - Devstral-Uncensored

**Popular model creators/builders**

- HauhauCS
- Huihui
- DavidAU
- SC117
- LuffyTheFox
- mradermacher
- AEON
- Heretic
- Genesis
- Abliterated

---

<div class="post-metadata">

**Author:** ![odirachukwu\_onyejefu](https://onehack.st/user_avatar/onehack.st/odirachukwu_onyejefu/32/43629_2.png) [@odirachukwu\_onyejefu](https://onehack.st/u/odirachukwu_onyejefu)\
**Post date:** [July 18, 2026, 11:51am UTC](https://onehack.st/t/every-uncensored-ai-model-for-any-pc-the-one-command-tool-to-break-any-model-yourself/323200/15 "2026-07-18T11:51:59Z")

</div>

> [@Edgar](#):
>
> Dolphin3-Cyber ​​8B

Thanks alot

---

<div class="post-metadata">

**Author:** ![GrimXLock](https://onehack.st/user_avatar/onehack.st/grimxlock/32/174321_2.png) [@GrimXLock](https://onehack.st/u/GrimXLock)\
**Post date:** [July 20, 2026, 2:55am UTC](https://onehack.st/t/every-uncensored-ai-model-for-any-pc-the-one-command-tool-to-break-any-model-yourself/323200/16 "2026-07-20T02:55:37Z")

</div>

I was inspired by abliterate repo and made my own tool  
[https://github.com/tjcrims0nx/annihilation-llm](https://github.com/tjcrims0nx/annihilation-llm) hybrid, moe, mulitmodels too. i just updated my tool and ported the features from Obliterus repo.  
,  
**OBLITERATUS Advanced Options**

You can now toggle experimental algorithms directly from the TUI configuration menu by selecting **OBLITERATUS Advanced**. This enables:

- **COSMIC Layer Selection** : Instead of blindly searching across the entire network, the system analyzes cosine similarities between harmless and harmful residual streams. It automatically anchors the optimization process around the mathematically proven optimal layer, massively reducing the search space.
- **Expert-Granular Abliteration (EGA)**: For Mixture-of-Experts (MoE) models, EGA calculates the alignment score of each expert’s weight matrix against the target refusal direction. Instead of applying a flat penalty, experts containing high concentrations of refusal vectors are aggressively modified while benign knowledge experts are perfectly preserved.
- **Gaussian-shaped Ablation Kernels** : Replaces traditional rigid interpolation bounds with a smooth, bell-shaped Gaussian curve to distribute weight changes across adjacent layers. This results in smoother vector blending and better text coherence post-ablation.

,