Uncensored

An instruction-tuned 24B model based on Mistral that minimizes refusals on sensitive or controversial topics.

Strengths

Handles requests without typical safety guardrails, giving responses on topics where many models decline or heavily restrict output.

Best for

Scenarios requiring unrestricted text generation on sensitive content, research into model behavior, or applications where the user takes responsibility for output filtering.

Limitations

The lack of safety guidelines makes it unsuitable for public-facing applications; outputs may contain harmful, misleading, or offensive content without moderation.