Venice Uncensored 1.1
A highly steerable, 24B-parameter uncensored model co-developed by Venice and Dolphin, running with end-to-end encryption.
Overview
What is Venice Uncensored 1.1
Venice Uncensored 1.1 is a 24B-parameter language model co-developed by Venice.ai and the Dolphin AI team (led by Eric Hartford). Released in July 2025, this model strips away artificial content restrictions while maintaining high conversational intelligence, offering a truly unrestricted and steerable AI experience.
Running it privately on Venice
On Venice, Venice Uncensored 1.1 runs with maximum sovereignty. It is deployed with end-to-end encryption and runs in a Trusted Execution Environment (TEE) under a strict zero-retention privacy tier. Because it has no hard-coded moralizing alignment, Venice users can set their own system prompts to define the model's behavior without fear of surveillance or judgment.
Assessment
Strengths and limitations
- Completely uncensored: Stripped of artificial moralizing refusals, making it ideal for creative writing, dark fiction, and unfiltered research.
- Highly steerable: Lacking hard-coded alignment, it relies entirely on your system prompt to set tone, perspective, and boundaries.
- Refined tone control: Version 1.1 is polite and neutral by default, but capable of adopting rude, explicit, or highly stylized personas when prompted.
- Improved ethical reasoning: Uses consequentialist reasoning to handle complex dilemmas with fewer biases and inconsistencies.
- End-to-end encrypted: Runs within Venice's secure TEE infrastructure, ensuring your prompts are never logged, stored, or visible to third parties.
- Smaller context window (32K tokens) compared to frontier models like Claude or DeepSeek.
- No native vision or multi-modal capabilities; limited to text-only inputs.
- Does not support native tool use or function calling, though it does support Venice's web search integration.
Capabilities
What it supports
- Tool use / function calling
- Vision (image input)
- Reasoning
- Web search
- Code-optimized
- Structured output (JSON schema)
- Audio input
- Video input
- Multiple image inputs
- Log probabilities
Specifications
Datasheet
- Maker
- Cognitive Computations & Venice.ai
- Released
- July 3, 2025
- Architecture
- Mistral 24B foundation (V7-Tekken chat template)
- Parameters
- 24 Billion
- Open weights
- Yes — Apache-2.0
- Context window
- 32K tokens
- Max output
- 4.096K tokens
- Capabilities
- Web search
- Privacy on Venice
- Private — zero retention
- Available on Venice since
- Mar 2026
API
Call it from your code
Venice exposes an OpenAI-compatible API. Point your base URL at Venice and pass the model id.
curl https://api.venice.ai/api/v1/chat/completions \
-H "Authorization: Bearer $VENICE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "e2ee-venice-uncensored-24b-p",
"messages": [{ "role": "user", "content": "Explain quantum tunneling simply." }]
}'Pricing
What it costs on Venice
Billed per token on Venice: $0.25 per 1M input tokens and $1.15 per 1M output tokens.
New Venice accounts include a free daily allowance and 500 welcome credits — no credit card required.
Alternatives
How it compares
| Model | Best for | Context window | Open weights | Uncensored design | Price (Venice) |
|---|---|---|---|---|---|
| Venice Uncensored 1.1 | The premier uncensored 24B model, running with end-to-end encryption. | 32K tokens | Yes | Yes (Zero refusals) | $0.25 / $1.15 |
| DeepSeek V3.2 | Much larger context and lower output pricing, but subject to standard safety filters. | 160K tokens | Yes | Standard alignment | $0.33 / $0.48 |
| Google Gemma 4 31B Instruct | Highly capable open-weights model, but heavily aligned and prone to refusals on sensitive topics. | 256K tokens | Yes | Strict alignment | $0.12 / $0.36 |
| Claude Sonnet 4.6 | Frontier intelligence and massive context, but proprietary and highly censored. | 1M tokens | No | Strict alignment | $3.60 / $18 |
The premier uncensored 24B model, running with end-to-end encryption.
Use cases
What it is good for
- 01Creative writing and screenwriting requiring unfiltered character dialogue, conflict, or mature themes.
- 02Academic and historical research into sensitive, controversial, or taboo topics without system refusals.
- 03Roleplay and interactive storytelling where the user defines the exact boundaries and character alignments.
- 04Drafting complex arguments or exploring philosophical dilemmas from multiple, non-standard perspectives.
Prompting
Getting better results
Be explicit in your system prompt. Since the model has no default alignment, it depends on your instructions to establish its persona.
Use the web search toggle on Venice when you need to ground its responses in real-time, up-to-date information.
If you want a specific tone (e.g., highly academic, sarcastic, or casual), state it clearly in the first turn.
Version history
Initial co-release stripping away artificial restrictions.
Current version with improved ethical reasoning, tone control, and creative writing.
FAQ
Frequently asked questions
Venice Uncensored 1.1 is a 24B-parameter language model built on Mistral's architecture and co-developed by Venice.ai and the Dolphin AI team. It is designed to be highly steerable, private, and completely free of artificial content restrictions.
On the Venice API, Venice Uncensored 1.1 is priced at $0.25 per 1M input tokens and $1.15 per 1M output tokens, making it an exceptionally cost-effective option for uncensored text generation.
Yes, Venice Uncensored 1.1 is an open-weights model released under the Apache-2.0 license. You can download the weights from Hugging Face or run it directly on Venice.
Unlike standard models (such as Claude or GPT) which have hard-coded moral alignments and frequently refuse prompts, Venice Uncensored 1.1 has no built-in censorship filters. It relies entirely on the user's system prompt to guide its behavior.
Yes, Venice Uncensored 1.1 supports web search on Venice, allowing you to ground its uncensored responses in real-time web data.
Claude Sonnet 4.6 is vastly superior for complex coding, reasoning, and large-document analysis due to its 1M context window. However, Claude is heavily censored and proprietary. Venice Uncensored 1.1 is the clear choice for unrestricted creative writing, sensitive research, and complete privacy.
Venice runs this model with end-to-end encryption in a Trusted Execution Environment (TEE). Venice has a zero-retention policy, meaning your prompts and generations are never stored, logged, or used to train models.
Run Venice Uncensored 1.1 privately
No prompt logging. No data used for training.