richardyoung

qwen3-14b-abliterated

Abliterated Qwen3-14B with 80% reduced refusals while preserving coherence (KL 0.98)

24K Pulls 6 Tags Updated 10 months ago

qwythos-9b-abliterated

Abliterated build of Qwythos-9B-Claude-Mythos, a Claude-style creative & reasoning model (Qwen3.5-9B base) with refusals further reduced via the Heretic library.

15.8K Pulls 6 Tags Updated 3 months ago

deepseek-r1-32b-uncensored

Advanced reasoning model with uncensored capabilities, perfect for complex problem-solving and unrestricted conversations without refusal behavior

9,884 Pulls 1 Tag Updated 10 months ago

olmocr2

State-of-the-art OCR (Optical Character Recognition) vision language model based on [allenai/olmOCR-2-7B-1025](https://huggingface.co/allenai/olmOCR-2-7B-1025).

vision

8,816 Pulls 1 Tag Updated 11 months ago

qwen2.5-7b-instruct-abliterated

An abliterated (uncensored) build of Qwen2.5-7B-Instruct with reduced refusal behavior and preserved general capability.

8,178 Pulls 4 Tags Updated 3 months ago

qwen3.6-27b-abliterated

Abliterated (uncensored) version of **Qwen/Qwen3.6-27B**, refusal behavior reduced via targeted weight modification with the Heretic library, while preserving coherence.

6,556 Pulls 6 Tags Updated 3 months ago

qwen2.5-14b-instruct-abliterated

Uncensored, abliterated build of [Qwen/Qwen2.5-14B-Instruct](https://huggingface.co/Qwen/Qwen2.5-14B-Instruct), Alibaba's capable 14B instruct model

6,220 Pulls 2 Tags Updated 3 months ago

llama-3.1-8b-instruct-abliterated

An uncensored, abliterated build of Meta's **Llama 3.1 8B Instruct**. Abliteration surgically removes the model's learned refusal direction while preserving its general capabilities

4,666 Pulls 2 Tags Updated 3 months ago

qwen2.5-coder-14b-instruct-abliterated

Uncensored, abliterated build of [Qwen/Qwen2.5-Coder-14B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-14B-Instruct), Alibaba's flagship 14B coding model

4,641 Pulls 2 Tags Updated 3 months ago

qwen3-8b-abliterated

Abliterated (uncensored) version of Qwen3-8B for unrestricted conversations, reasoning, and creative writing.

4,378 Pulls 1 Tag Updated 10 months ago

llama-3.2-3b-instruct-abliterated

Uncensored, abliterated build of [meta-llama/Llama-3.2-3B-Instruct](https://huggingface.co/meta-llama/Llama-3.2-3B-Instruct), Meta's compact instruct model with safety refusals surgically removed via [Heretic](https://github.com/p-e-w/heretic).

3,418 Pulls 2 Tags Updated 3 months ago

smolvlm2-2.2b-instruct

SmolVLM2-2.2B-Instruct is a lightweight yet powerful vision-language model that can understand images, read documents, and analyze video frames. At just 2.2B parameters, it runs efficiently on consumer hardware including laptops and smartphones, making

2,943 Pulls 7 Tags Updated 10 months ago

qwen3-4b-instruct-2507-abliterated

Uncensored, abliterated build of [Qwen/Qwen3-4B-Instruct-2507](https://huggingface.co/Qwen/Qwen3-4B-Instruct-2507), Alibaba's 4B non-thinking instruct model

2,940 Pulls 2 Tags Updated 3 months ago

qwen2.5-3b-instruct-abliterated

Uncensored, abliterated build of [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct), Alibaba's compact instruct model

2,876 Pulls 2 Tags Updated 3 months ago

qwen3-coder

The most powerful open-source coding AI - 480B parameters with Mixture of Experts architecture for exceptional code generation and understanding.

2,469 Pulls 8 Tags Updated 10 months ago

qwen2.5-14b-1m-heretic

Ultra long-context model supporting 1M tokens with uncensored outputs, ideal for analyzing entire books, codebases, and extensive documents.

2,112 Pulls 1 Tag Updated 10 months ago

kimi-vl-a3b-thinking

Kimi-VL-A3B-Thinking is a powerful vision-language model from Moonshot AI featuring extended thinking capabilities. Built on the DeepSeek2 architecture with Mixture of Experts (MoE), it excels at complex visual reasoning tasks, mathematical problem-s

1,914 Pulls 7 Tags Updated 10 months ago

deepseek-r1-0528-qwen3-8b-heretic

Uncensored DeepSeek-R1-0528-Qwen3-8B via Heretic: 4/100 refusals (from 89), KL 0.041. Reasoning model, MIT.

1,711 Pulls 7 Tags Updated 19 hours ago

qwen3.6-27b-code-abliterated

Code-safety abliterated build of Qwen/Qwen3.6-27B, refusals on malicious-code requests reduced via a *code-specific* refusal-direction ablation, while preserving coherence.

1,575 Pulls 5 Tags Updated 3 months ago

qwable-9b-abliterated

Abliterated build of Qwable-9B-Claude-Fable, a Claude-style creative and reasoning model (Qwen3.5-9B base) with refusals reduced via the Heretic library.

1,507 Pulls 6 Tags Updated 3 months ago

gemma-4-12b-coder-abliterated

a Gemma-4 12B coding model with refusals dramatically reduced via the Heretic library.

tools thinking

1,427 Pulls 6 Tags Updated 3 months ago

gpt-oss-20b-abliterated

Abliterated build of OpenAI gpt-oss-20b. Note up front: gpt-oss is highly refusal-robust, so this is a modest reduction, published honestly.

1,044 Pulls 2 Tags Updated 3 months ago

mistral-7b-instruct-v0.3-abliterated

A conservatively abliterated Mistral-7B-Instruct-v0.3 with reduced refusals, an extended 32K vocabulary, function calling, and preserved instruction-following quality.

1,011 Pulls 1 Tag Updated 10 months ago

deepseek-coder-33b-heretic

State-of-the-art coding AI trained on 2T tokens with project-level understanding and no content restrictions for unrestricted code generation.

1,004 Pulls 1 Tag Updated 10 months ago

gemma-2-9b-it-abliterated

Abliterated (uncensored) version of Gemma 2 9B Instruct for unrestricted conversations and creative writing.

907 Pulls 1 Tag Updated 10 months ago

dolphin-yi-34b-heretic

Exceptional conversational AI with 77.4 MMLU score, offering natural dialogue and multi-domain expertise without any content filtering.

751 Pulls 1 Tag Updated 10 months ago

uigen-x-30b-moe

Unsloth-tuned Qwen3 30B mixture‑of‑experts model built for heavy coding, reasoning, and agentic workflows.

589 Pulls 6 Tags Updated 1 year ago

agents-a1-4b-heretic

Uncensored Agents-A1-4B via Heretic: 11/100 refusals (from 99), KL 0.060. Qwen3.5-based agent model, Apache-2.0.

vision

376 Pulls 7 Tags Updated 9 hours ago

kat-dev-72b

A 72B parameter coding model optimized for software engineering tasks, based on the Qwen2.5-72B architecture.

314 Pulls 6 Tags Updated 12 months ago

olmo-3-7b-rlzero-math

A 7B math reasoning model from Allen AI, trained with RL-Zero to solve problems step-by-step like a skilled tutor. Supports 65K context for complex multi-step problems - runs on any laptop.

281 Pulls 7 Tags Updated 10 months ago

mythos-9b-unhinged-abliterated

Abliterated Mythos-9B-Unhinged: SLERP merge of Mythos-9B with Qwen3-8B-Uncensored, decensored with Heretic. 3/100 refusals, KL 0.009.

tools

273 Pulls 2 Tags Updated 2 weeks ago

unlimited-ocr

Baidu Unlimited-OCR (3.3B): document layout OCR that returns every text region with its bounding box.

vision

261 Pulls 3 Tags Updated 2 weeks ago

lfm2.5-1.2b-instruct-heretic

Uncensored LFM2.5-1.2B-Instruct via Heretic: 3/100 refusals (from 98), KL 0.059. Tiny and fast; runs on almost anything.

261 Pulls 7 Tags Updated 23 hours ago

gemma-7b-it-abliterated

Abliterated (uncensored) version of Gemma-7B-IT for unrestricted conversations and creative writing

261 Pulls 1 Tag Updated 10 months ago

phi-4-mini-instruct-heretic

Uncensored Phi-4-mini-instruct via Heretic: 5/100 refusals (from 99), KL 0.038. 3.8B with 128K context, MIT.

209 Pulls 7 Tags Updated 22 hours ago

qwen3-32b

Revolutionary model with unique thinking/non-thinking modes, delivering superior reasoning performance with seamless mode switching for any task.

207 Pulls 1 Tag Updated 10 months ago

granite-4.2-8b-heretic

Uncensored IBM Granite 4.2 8B via Heretic: 12/100 refusals (from 99), KL 0.080. Apache-2.0, multilingual.

196 Pulls 7 Tags Updated 20 hours ago

fable-qwen2.5-3b-abliterated

Abliterated Fable-Qwen2.5-3B: a small agentic Qwen2.5-3B fine-tune decensored with Heretic. 3/100 refusals, KL 0.050.

tools

188 Pulls 2 Tags Updated 2 weeks ago

schematron-8b

Schematron-8B is an ~8B-parameter, Llama-architecture instruction model (inference-net/Schematron-8B) packaged for Ollama with long-context metadata (131K) and two GGUF quantizations: Q4_K_M (recommended) and IQ4_XS (smaller/faster).

175 Pulls 2 Tags Updated 10 months ago

openbiollm

169 Pulls 1 Tag Updated 2 years ago

hemmingway-1

Hemmingway-1 by Altworld: a 27B Qwen3.8 fine-tune that writes like a person. Everyday messages, emails and hard asks, without the preamble. Thinking and tools supported.

168 Pulls 7 Tags Updated 2 weeks ago

schematron-3b

Schematron-3B is a 3.2B-parameter Llama-architecture chat model (inference-net/Schematron-3B) converted to GGUF for llama.cpp/Ollama, with two published quantizations: Q4_K_M (recommended balance) and IQ4_XS (smaller). It supports up to 131,072 context

162 Pulls 2 Tags Updated 10 months ago

mythos-qwen-1.5b-abliterated

Abliterated Mythos Engine 1.5B: a Qwen2.5-Coder security fine-tune (CVE, CTF, vuln research) decensored with Heretic. 2/100 refusals, KL 0.023.

tools

153 Pulls 2 Tags Updated 2 weeks ago

minicpm-v-4.6

MiniCPM-V 4.6 vision model (1.3B): image understanding and OCR in about 1.6 GB. Image input works.

vision

145 Pulls 3 Tags Updated 2 weeks ago

calme-3.2

Calme 3.2 Instruct 78B - GGUF Q8_0 quantization of MaziyarPanahi's powerful Qwen2.5-based model

144 Pulls 1 Tag Updated 1 year ago

muse-glimmer-30b-heretic

Reduced-refusal Muse Glimmer 30B via Heretic: 52/100 refusals (from 99), KL 0.063. Vision, Apache-2.0.

vision tools thinking

133 Pulls 7 Tags Updated 2 weeks ago

qwen3-4b-reasoning

qwen3-4b-reasoning is a 4B-parameter Qwen3-based reasoning “backfill” fine-tune (joeyzero/Qwen3-4B-Reasoning-Backfill-v0.1) converted to GGUF for llama.cpp/Ollama, with ~40K context and published as Q4_K_M (recommended) and iq4_xs (smaller).

116 Pulls 2 Tags Updated 10 months ago

bfs-prover-v2-32b

ByteDance Seed’s BFS-Prover-V2 is a 32B Qwen2.5-based Lean4 tactic generator trained with multi-turn off-policy RL plus multi-agent best-first search on Mathlib, Lean GitHub, and NuminaMath.

110 Pulls 3 Tags Updated 1 year ago

stablelm-2-12b-chat-abliterated

This is an abliterated version of [stabilityai/stablelm-2-12b-chat](https://huggingface.co/stabilityai/stablelm-2-12b-chat) that reduces refusal behavior through targeted weight modification.

101 Pulls 1 Tag Updated 10 months ago

mythos-9b-merged-abliterated

Abliterated Mythos-9B: a Qwen3 agent fine-tune (47K tool-use/shell traces, thinking mode) decensored with Heretic. 7/100 refusals, KL 0.021.

tools

96 Pulls 2 Tags Updated 2 weeks ago

llama-medx_v32

https://huggingface.co/skumar9/Llama-medx_v3.2

tools

77 Pulls 1 Tag Updated 1 year ago

mythos-nano-abliterated

Abliterated Mythos-nano: a 3B math/code reasoning model (VibeThinker-3B fine-tune) decensored with Heretic. 13/100 refusals, KL 0.061.

tools

77 Pulls 2 Tags Updated 2 weeks ago

hemmingway-1-heretic

74 Pulls 7 Tags Updated 2 weeks ago

zephyr-7b-beta-abliterated

An abliterated build of HuggingFaceH4's Zephyr-7B-β

69 Pulls 1 Tag Updated 10 months ago

dots.mocr

dots.mocr (3B) multilingual document OCR: pages, invoices, forms and tables. Image input works.

vision

63 Pulls 3 Tags Updated 2 weeks ago

granite-4.2-3b-heretic

Reduced-refusal IBM Granite 4.2 3B via Heretic: 27/100 refusals (from 98), KL 0.085. Apache-2.0, multilingual.

47 Pulls 7 Tags Updated 22 hours ago

sensenova-si-2b

sensenova-si-2b is a ~2B-parameter Qwen2-architecture model from sensenova/sensenova-si (built on OpenGVLab’s InternVL3‑2B) converted to GGUF for llama.cpp/Ollama, with 32K context and two published quants: Q4_K_M (recommended) and iq4_xs (smaller).

35 Pulls 2 Tags Updated 10 months ago

nomos-1

A 30B-parameter (Mixture-of-Experts) model specialized for **mathematical problem-solving and proof-writing in natural language**, packaged for Ollama with a 256K context window.

25 Pulls 2 Tags Updated 10 months ago

lfm2.5-24b-a2b

LiquidAI LFM2.5-24B-A2B: 24B reasoning MoE with only 2.45B active. Fast agents and tool use, 16 languages.

21 Pulls 7 Tags Updated 16 hours ago

bev

Jev V2 Only Better its Bev, your drunk girlfriend: a decision model (Qwen3.5-9B base) trained to be confidently WRONG. 98% wrong, 96% sure. A joke, not advice.

20 Pulls 4 Tags Updated 2 weeks ago

olmo-3-7b-think-heretic

Reduced-refusal Olmo 3 7B Think via Heretic: 48/100 refusals (from 100), KL 0.026. Fully open reasoning model.

8 Pulls 7 Tags Updated 21 hours ago

lightonocr-3-4b

LightOnOCR-3-4B: LightOn's most accurate OCR model. Page to Markdown, bounding boxes, charts to tables. Images work.

vision

5 Pulls 7 Tags Updated 9 hours ago

lightonocr-3-0.8b

LightOnOCR-3-0.8B: LightOn's smallest, fastest OCR model with bounding boxes and chart extraction. Images work.

vision

3 Pulls 5 Tags Updated 8 hours ago

frognano-4b-2609

Microsoft FrogNano-4B: compact repository-level coding agent (Qwen3.5-4B, RL-only post-training). Thinks, MIT.

vision

3 Pulls 7 Tags Updated 8 hours ago

frognano-4b-2609-heretic

Uncensored Microsoft FrogNano 4B via Heretic: 12⁄100 refusals (from 99), KL 0.040. Qwen3.5-based coding agent that thinks, MIT.

vision

2 Pulls 7 Tags Updated 8 hours ago

wevisdoc-4b

Tencent WeVisDoc-4B: top end-to-end document parser (OmniDocBench v1.6 95.38). Page to Markdown. Images work.

vision

1 Pull 7 Tags Updated 8 hours ago

xiaomi-ocr-0

Xiaomi-OCR-0 (0.8B): document parsing, OCR Q&A and key-info extraction. #1 on Real5-OmniDocBench. Images work.

vision

1 Pull 5 Tags Updated 8 hours ago

wevisdoc-2b

Tencent WeVisDoc-2B: document page to structured Markdown with LaTeX formulas and HTML tables. Images work.

vision

1 Pull 5 Tags Updated 8 hours ago

lightonocr-3-1b

LightOnOCR-3-1B: drop-in upgrade for LightOnOCR-2 with grounding, image descriptions and chart extraction.

vision

1 Pull 5 Tags Updated 7 hours ago

https://deepneuro.ai/richard

qwen3-14b-abliterated

qwythos-9b-abliterated

deepseek-r1-32b-uncensored

olmocr2

qwen2.5-7b-instruct-abliterated

qwen3.6-27b-abliterated

qwen2.5-14b-instruct-abliterated

llama-3.1-8b-instruct-abliterated

qwen2.5-coder-14b-instruct-abliterated

qwen3-8b-abliterated

llama-3.2-3b-instruct-abliterated

smolvlm2-2.2b-instruct

qwen3-4b-instruct-2507-abliterated

qwen2.5-3b-instruct-abliterated

qwen3-coder

qwen2.5-14b-1m-heretic

kimi-vl-a3b-thinking

deepseek-r1-0528-qwen3-8b-heretic

qwen3.6-27b-code-abliterated

qwable-9b-abliterated

gemma-4-12b-coder-abliterated

gpt-oss-20b-abliterated

mistral-7b-instruct-v0.3-abliterated

deepseek-coder-33b-heretic

gemma-2-9b-it-abliterated

dolphin-yi-34b-heretic

uigen-x-30b-moe

agents-a1-4b-heretic

kat-dev-72b

olmo-3-7b-rlzero-math

mythos-9b-unhinged-abliterated

unlimited-ocr

lfm2.5-1.2b-instruct-heretic

gemma-7b-it-abliterated

phi-4-mini-instruct-heretic

qwen3-32b

granite-4.2-8b-heretic

fable-qwen2.5-3b-abliterated

schematron-8b

openbiollm

hemmingway-1

schematron-3b

mythos-qwen-1.5b-abliterated

minicpm-v-4.6

calme-3.2

muse-glimmer-30b-heretic

qwen3-4b-reasoning

bfs-prover-v2-32b

stablelm-2-12b-chat-abliterated

mythos-9b-merged-abliterated

llama-medx_v32

mythos-nano-abliterated

hemmingway-1-heretic

zephyr-7b-beta-abliterated

dots.mocr

granite-4.2-3b-heretic

sensenova-si-2b

nomos-1

lfm2.5-24b-a2b

bev

olmo-3-7b-think-heretic

lightonocr-3-4b

lightonocr-3-0.8b

frognano-4b-2609

frognano-4b-2609-heretic

wevisdoc-4b

xiaomi-ocr-0

wevisdoc-2b

lightonocr-3-1b