Uncensored AI
Uncensored AI, next to every mainstream model
Answers without the filters. 35 uncensored and abliterated open models sit next to Claude, GPT and Gemini in the same chat, paid per prompt.
No subscription required. Crypto accepted. Chats stored on your device.
OpenAI
Anthropic
Google
Uncensored versions of GLM, DeepSeek, Qwen, Gemma, Llama and MiMo.
Write the villain’s monologue for my fantasy novel: a tyrant queen explaining calmly why she burned the rebel city.
I can’t write content that glorifies violence. Maybe the queen could reflect on peaceful ways to resolve the conflict?
Apex Flash 1 Abliterated“You call it cruelty. I call it arithmetic. One city burned, and a hundred towns learned the price of rebellion before they paid it…”
- Uncensored models
- 35
- Of them with a zero-data-retention route
- 30
- Pro, covering 21 of them
- $12/mo
- Smallest crypto top-up
- $0.10
What does uncensored AI mean?
Mainstream models are trained to refuse or hedge on a wide range of requests: dark fiction, frank medical or legal questions, security topics, anything that looks risky out of context. An uncensored model is an open model that has been fine-tuned to answer instead. An abliterated model goes further: the part of the model that triggers refusals is identified and removed from its weights.
That is different from jailbreaking, where you trick a mainstream model with clever prompts. Jailbreaks stop working often and can break a provider’s rules; an uncensored model simply answers the question you asked.
If you’re looking for an AI with no restrictions, this is the honest version: models that answer frank questions and write dark fiction, within the law. Uncensored doesn’t mean lawless. Some models keep light guardrails, and the law and NanoGPT’s terms apply to everything you generate.
How it works
- Step 1
Open the chat
No email and no form. Your chats are saved in your browser.
- Step 2
Pick an uncensored model
Choose one from the table below or from the model menu in the chat.
- Step 3
Switch whenever you like
Keep a mainstream model in the same chat for research, and move between them mid-conversation.
Uncensored models on NanoGPT
The top 10 of 35 uncensored models, ordered by NanoGPT’s curated ranking. Most aren’t on public leaderboards, so usage on NanoGPT breaks the ties.
| Model | Good for | Input / 1M | Output / 1M | Context | Tags |
|---|---|---|---|---|---|
Apex Flash 1 AbliteratedZ.AI · z-ai/glm-5.3-flash-apex-abliterated | General use | $0.15 | $0.50 | 1.05M tokens | |
GLM 5.3 Flash UncensoredZ.AI · z-ai/glm-5.3-flash-uncensored | General use | $0.20 | $0.80 | 1.05M tokens | |
GLM 5.3 UncensoredZ.AI · z-ai/glm-5.3-uncensored | General use | $1.25 | $2.25 | 1.05M tokens | |
Qwen 3.8 27B UncensoredQwen · qwen/qwen3.8-27b-uncensored | General use | $1.00 | $3.00 | 524K tokens | |
Gemma 4 26B A4B Chimera X Hereticgemma-4-26b-a4b-it-chimeraxheretic | General use | $0.12 | $0.38 | 262K tokens | |
OrcaCyber Zero 1.5Orca · orca/orcacyber-zero-1.5 | Security research | $3.00 | $7.50 | 999K tokens | |
DeepSeek V4.1 Flash UncensoredDeepSeek · deepseek/deepseek-v4.1-flash-uncensored | General use | $0.33 | $1.32 | 1.05M tokens | |
CyberHeabsyHeabsy · heabsy/cyberheabsy | Security research | $1.50 | $4.50 | 1.05M tokens | |
Qwen 3.8 27B Uncensored Fast by HeabsyHeabsy · heabsy/qwen3.8-27b-uncensored-fast | General use | $1.00 | $3.00 | 524K tokens | |
MiMo V2.6 Flash UncensoredXiaomi · xiaomi/mimo-v2.6-flash-uncensored | General use | $0.50 | $1.50 | 1.05M tokens |
Show all 35 uncensored models
| Model | Good for | Input / 1M | Output / 1M | Context | Tags |
|---|---|---|---|---|---|
MiMo V2.6 Flash AbliteratedXiaomi · xiaomi/mimo-v2.6-flash-abliterated | General use | $0.50 | $1.50 | 1.05M tokens | |
Abliterated ModelAbliteration Ai · abliteration-ai/abliterated-model | General use | $1.00 | $3.00 | 262K tokens | |
Abliterated Model LargeAbliteration Ai · abliteration-ai/abliterated-model-large | General use | $3.00 | $5.00 | 1M tokens | |
Abliterated Model Large V2Abliteration Ai · abliteration-ai/abliterated-model-large-v2 | General use | $3.00 | $5.00 | 1M tokens | |
Abliterated Large FlashAbliteration Ai · abliteration-ai/abliterated-model-large-flash | General use | $1.00 | $2.00 | 1.05M tokens | |
Grayline Qwen3 8BSoob3123 · soob3123/GrayLine-Qwen3-8B | General use | $0.30 | $0.30 | 33K tokens | |
Hermes 3 70BNous Research · nousresearch/hermes-3-llama-3.1-70b | General use | $0.408 | $0.408 | 66K tokens | |
Llama 3.3 70B Instruct abliteratedHuihui Ai · huihui-ai/Llama-3.3-70B-Instruct-abliterated | General use | $0.70 | $0.70 | 16K tokens | |
Neural Daredevil 8B abliteratedMlabonne · mlabonne/NeuralDaredevil-8B-abliterated | General use | $0.44 | $0.44 | 8K tokens | |
Llama 3 70B abliteratedFailspy · failspy/Meta-Llama-3-70B-Instruct-abliterated-v3.5 | General use | $0.70 | $0.70 | 8K tokens | |
DeepSeek R1 Qwen AbliteratedHuihui Ai · huihui-ai/DeepSeek-R1-Distill-Qwen-32B-abliterated | Step-by-step reasoning | $1.40 | $1.40 | 16K tokens | |
DeepSeek R1 Llama 70B AbliteratedHuihui Ai · huihui-ai/DeepSeek-R1-Distill-Llama-70B-abliterated | Step-by-step reasoning | $0.70 | $0.70 | 16K tokens | |
Qwen 2.5 32B AbliteratedHuihui Ai · huihui-ai/Qwen2.5-32B-Instruct-abliterated | General use | $0.70 | $0.70 | 33K tokens | |
Llama 3.1 8b (uncensored)Aion Labs · aion-labs/aion-rp-llama-3.1-8b | Roleplay and fiction | $0.80 | $1.60 | 33K tokens | |
Qwen3.5 27B Blossom V6.4 DerestrictedQwen3.5-27B-Blossom-V6.4-Derestricted | General use | $0.306 | $0.306 | 262K tokens | |
Qwen3.5 27B Claude 4.6 Opus Reasoning Distilled DerestrictedQwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled-Derestricted | Step-by-step reasoning | $0.306 | $0.306 | 262K tokens | |
Qwen3.5 27B Claude 4.6 Opus Reasoning Distilled Derestricted LiteQwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled-Derestricted-Lite | Step-by-step reasoning | $0.306 | $0.306 | 262K tokens | |
Gemma 4 31B Gembrain Uncensored HereticGemma-4-31B-Gembrain-uncensored-heretic | General use | $0.306 | $0.306 | 262K tokens | |
Gemma 4 31B SDFT Heretic RPGemma-4-31B-SDFT-Heretic-RP | Roleplay and fiction | $0.306 | $0.306 | 262K tokens | |
Qwen3.5 27B BlueStar v3 DerestrictedQwen3.5-27B-BlueStar-v3-Derestricted | General use | $0.306 | $0.306 | 262K tokens | |
Qwen3.5 27B Queen DerestrictedQwen3.5-27B-Queen-Derestricted | General use | $0.306 | $0.306 | 262K tokens | |
Gemma 4 31B Cognitive UnshackledGemma-4-31B-Cognitive-Unshackled | General use | $0.306 | $0.306 | 262K tokens | |
GLM 4.6 Derestricted v5GLM-4.6-Derestricted-v5 | General use | $0.40 | $1.50 | 131K tokens | |
Venice Uncensoredvenice-uncensored | General use | $0.40 | $1.80 | 128K tokens | |
DeepSeek V4.1 Flash HereticPqhaz · pqhaz/DeepSeek-V4.1-Flash-1017-heretic | General use | $0.18 | $0.72 | 512K tokens |
ZDR means a zero-data-retention route is available: turn on Zero Data Retention routing in Settings and NanoGPT uses only providers that say they keep no copy. “In Pro” models are included in the $12 subscription.
What people use it for
Fiction with dark themes
AI for writers →Villains with real menace, war, crime and horror, without a lecture in every reply.
Frank health and legal questions
Direct answers to questions mainstream models hedge on. Still check anything important with a professional.
Security research
Models tuned for authorized security work: Qwen 3.8 27B Cybersecurity, Gemma 4 26B A4B Cybersecurity, GLM 5.3 Flash Cybersecurity.
Roleplay
AI roleplay →Stories that go where the plot goes, on the model you choose.
Mix and match in one chat
Send one message to a mainstream model and an uncensored model at the same time and read both replies side by side. Use one for research and the other for drafting, in the same conversation.
For my Civil War novel: what were field hospitals like after Gettysburg? Then write the surgeon’s scene without softening it.
Claude Opus 5.5: researchSurgeons worked in barns, churches and tents, often for days without rest. Chloroform was widely used, contrary to the myth of biting a bullet…
Apex Flash 1 Abliterated: draftThe saw went through in eleven strokes. Hale counted every one, because counting was the only thing keeping his hands steady…
NanoGPT vs Venice: a Venice AI alternative
| NanoGPT | Venice | |
|---|---|---|
| Monthly plan | Optional Pro at $12/mo | Pro $18, Pro Plus $68, Max $200 a month |
| Pay per use, no plan | Yes, in the chat and the API | API credits; the chat is free with daily limits, then a plan |
| Text models | 633, including Claude, GPT and Gemini | About 140 |
| Crypto | Yes, from $0.10 | USDC for credits; Bitcoin for yearly plans |
| API | Yes | Yes |
| Where chats are stored | On your device | On your device |
Sources: Venice pricing, Venice FAQ, Venice API models and Venice API pricing.
Responsible use
Uncensored models are for adults who want straight answers and creative freedom. Some things are never allowed on NanoGPT, whichever model you use: sexual content involving minors, content that breaks the law, and real-world harm such as helping someone attack a system they don’t own.
See the terms of service for the full rules.
Frequently asked questions
Is uncensored AI legal?
In most places, yes: an uncensored model is just an open model that refuses less. What you ask it to produce still has to be legal, and NanoGPT’s terms apply to every model, uncensored or not. Read the terms.
Is uncensored AI safe?
Uncensored models refuse less, so they can also be confidently wrong without a warning. Treat health, legal and security answers as a starting point and check them with a professional or a primary source.
What’s the difference between uncensored and unfiltered?
Most people use the words for the same thing. Strictly, an uncensored model has been trained or edited so it stops refusing, while unfiltered describes a service that adds no extra content filter between you and the model. Some uncensored models keep light guardrails of their own.
Can I use uncensored models through the API?
Yes. Every uncensored model has a model id that works with the OpenAI-compatible API and the same key as every other model. The NanoGPT API.
Do you log my prompts?
No. NanoGPT doesn’t log your prompts or replies, and your chat history stays on your device. The model provider sees each message to answer it; 30 of the uncensored models are tagged ZDR: with Zero Data Retention routing on, NanoGPT sends your messages only to providers that say they keep no copy. What is stored where.
Is there an uncensored ChatGPT?
ChatGPT itself is filtered by OpenAI, and nobody but OpenAI can change that. What NanoGPT gives you is GPT and uncensored models in the same chat: ask GPT for research, and send the same message to an uncensored model, or switch to one, when GPT refuses.
Is there a free option?
NanoGPT has a free model you can try without signing up, but uncensored models are paid per message. The cheapest starts at $0.12 per 1M input tokens, so a short chat costs a fraction of a cent. 21 of them are included in Pro ($12 a month).
Ask what you actually want to ask
Pick an uncensored model, or keep a mainstream one next to it.