Venice: Uncensored is a Dolphin Mistral 24B fine-tune that traces its lineage to Mistral-Small-24B-Instruct-2501, developed through a collaboration between dphn.ai and Venice.ai. The "uncensored" framing is the central design choice: default safety and alignment layers are stripped away so that system prompts, behavioral constraints, and content boundaries remain fully under the operator's control. The model's open weights are hosted on Hugging Face under cognitivecomputations/Dolphin-Mistral-24B-Venice-Edition, giving technical users direct access to the artifact itself rather than only an API surface.
In practical terms, the model is aimed at advanced and unrestricted use cases where steerability and transparent behavior matter more than guardrails: red-teaming, policy gap analysis, controversial or edgy creative writing, and prototyping prompts that would be refused by mainstream assistants. Directory coverage describes it as a long-context text model suited to handling lengthy documents or multi-turn conversations without truncation, and it sits in a budget pricing tier that makes high-volume experimentation affordable. It is best treated as a specialized tool rather than a general-purpose assistant: useful for sensitive-topic summarization, brainstorming, and safety research where fewer content restrictions are an asset, and worth pairing with a higher-accuracy model for critical or customer-facing work.
