DeepSeek R1 Uncensored
Community · March 2025
● activeOpen Weightmixture of expertstext
Parameters671B (37B active)
Context Window128K tokens
Description
Community-modified version of DeepSeek R1 with built-in content restrictions removed using abliteration (a mathematical technique that surgically removes refusal behavior). Combines DeepSeek R1's powerful chain-of-thought reasoning — where the model 'thinks step by step' before answering — with unrestricted output, making it popular for coding and math tasks on consumer hardware.
Notable Milestones
- ▸Unrestricted coding assistant without content filters
- ▸Mathematical reasoning without safety-related refusals
Key Innovations
Abliteration
AbliterationRemoving safety guardrails from a model through targeted fine-tuning or weight manipulation. Controversial but popular in open-source community.
Reasoning
ReasoningStructured step-by-step problem solving, often using chain-of-thought or tree-of-thought approaches.
Open Weight
Open WeightModel weights are publicly released but training data/code may not be. Enables fine-tuning but not full reproduction.
Family Tree
Built On
Lineage
Related Research (1)
DeepSeek-R1Reasoning
2025 · DeepSeek AI
Demonstrated that pure RL training (without supervised fine-tuning on reasoning traces) can produce chain-of-thought reasoning, achieving performance …
More from Community / Uncensored
WizardLM2023-06 · 13B
Dolphin (Eric Hartford)2023-07 · —
Hermes (Nous Research)2023-09 · —
LLaMA 4 Scout Abliterated2025-05 · 81B (17B active × 16 experts)
Bark2023-04 · —
Mistral 7B Uncensored2024-01 · 7B
Qwen 2.5 72B Abliterated2025-01 · 72B
SOLAR 10.7B Uncensored2024-02 · 10.7B
LLaMA 3.1 405B Abliterated2024-08 · 405B
Gemma 2 Abliterated2024-08 · 27B