Venice Uncensored 1.2
Venice's flagship 24B uncensored model — built on Mistral architecture with native vision, web search, and zero refusal behavior.
Overview
What is Venice Uncensored 1.2
Venice Uncensored 1.2 is a 24-billion-parameter open-weights model designed for maximum creative freedom, unfiltered dialogue, and complex roleplay. Released in April 2026, it features a 128K context window, native vision capabilities, and tool use, built on a Dolphin-tuned Mistral architecture.
Running it privately on Venice
Running Venice Uncensored 1.2 on Venice guarantees absolute privacy with zero retention on your prompts. Unlike centralized platforms that log and monitor your conversations, Venice processes your requests permissionless-ly under a strict private tier. It is the ultimate tool for unbiased exploration, roleplay, and sovereign analysis without Big Tech surveillance.
Assessment
Strengths and limitations
- Carries the 'most uncensored' designation on Venice, offering zero refusal behavior for roleplay, creative writing, and unfiltered research.
- Native vision support allowing multi-image inputs and visual reasoning alongside text completions.
- Trained on decentralized compute (Bittensor's Targon subnet) using KTO reinforcement learning to preserve raw model capabilities without abliteration.
- Equipped with web search and tool use / function calling for real-time grounding and structured JSON outputs.
- Generous 128K context window allowing deep, long-form conversational memory.
- At 24B parameters, it may lack the deep reasoning of massive closed models like Claude Opus on highly complex coding tasks.
- Uncensored nature means it lacks safety guardrails; users must self-police output appropriateness.
- Served at fp16 precision which requires substantial compute, though optimized on Venice's decentralized network.
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
- Venice AI & Dolphin AI
- Released
- April 1, 2026
- Architecture
- Mistral 24B (Dolphin-Mistral-24B)
- Parameters
- 24 Billion
- Precision
- fp16
- Open weights
- Yes (Apache-2.0)
- Context window
- 128K tokens
- Max output
- 8.192K tokens
- Capabilities
- Vision, Function calling, Web search
- Privacy on Venice
- Private — zero retention
- Available on Venice since
- Apr 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": "venice-uncensored-1-2",
"messages": [{ "role": "user", "content": "Explain quantum tunneling simply." }]
}'Pricing
What it costs on Venice
Billed per token on Venice: $0.20 per 1M input tokens and $0.90 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 | Context window | Strongest at | Open weights | Price (Venice) |
|---|---|---|---|---|
| Venice Uncensored 1.2 | 128K tokens | Uncensored roleplay & vision | Yes | $0.20 / $0.90 |
| DeepSeek V3.2 | 160K tokens | Coding & general reasoning | Yes | $0.33 / $0.48 |
| Google Gemma 4 31B Instruct | 256K tokens | Instruction following | Yes | $0.12 / $0.36 |
| Claude Sonnet 4.6 | 1M tokens | Enterprise reasoning & coding | No | $3.60 / $18 |
The ultimate uncensored model with vision and zero retention.
Use cases
What it is good for
- 01Unrestricted creative writing, world-building, and complex character roleplay.
- 02Unbiased research on controversial, sensitive, or highly regulated topics.
- 03Visual analysis of documents, charts, or scenes without puritanical content filters.
- 04Private, local-first workflows requiring structured JSON outputs and tool integration.
Prompting
Getting better results
Speak naturally and directly; there is no need to 'jailbreak' or use complex framing to bypass safety filters.
Utilize the 128K context by feeding in extensive lore, character sheets, or long documents for analysis.
For structured tasks, explicitly request JSON schema outputs to leverage its native function-calling capabilities.
Version history
Predecessor 24B model.
Current flagship with vision, web search, and 128K context.
FAQ
Frequently asked questions
Venice Uncensored 1.2 is a flagship 24-billion-parameter open-weights text and vision model. Developed in collaboration with Dolphin AI and trained on Bittensor's Targon subnet, it is optimized for unfiltered dialogue, roleplay, and creative freedom without Big Tech censorship.
Venice Uncensored 1.2 is highly economical, costing $0.20 per 1 million input tokens and $0.90 per 1 million output tokens.
Yes, Venice Uncensored 1.2 is an open-weights model distributed under the Apache-2.0 license. You can try it on Venice's free tier or access it via the paid API.
Yes. Unlike many uncensored models, Venice Uncensored 1.2 natively supports vision (including multiple image inputs), web search, structured JSON outputs, and tool use / function calling.
Choose Venice Uncensored 1.2 if you require absolute creative freedom, NSFW/unfiltered roleplay, or zero-refusal behavior. Choose DeepSeek V3.2 for heavy coding or standard academic tasks where guardrails are not an issue.
Venice runs Venice Uncensored 1.2 under a strict private tier with zero retention. Your prompts and generations are never stored, logged, or used to train future models, ensuring complete sovereignty over your data.
The model features a 128K-token context window, allowing you to process long documents, extensive chat histories, or detailed roleplay scenarios.
Run Venice Uncensored 1.2 privately
No prompt logging. No data used for training.