Skip to content
TheChatArchiveThe Chat Archive
gpt-4o-search-preview· OpenAIHugging Face LMArena Search Arena 24K

Any uncensored local AIs out there I can download? Preferably released in 2025.

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with Geminigemini-2.5-pro-exp-03-25-grounding’s answer.
  1. User

    Any uncensored local AIs out there I can download? Preferably released in 2025.

  2. Assistant

    As of May 2025, several uncensored AI models have been released that you can download and run locally. Here are some notable options:

    Dolphin 3
    Developed by Cognitive Computations, Dolphin 3 is based on the Mistral 24B model. It offers unrestricted responses with no content filtering, making it suitable for logic and reasoning tasks. Note that it requires high-end hardware for optimal performance.

    Hermes 3
    Created by Nous Research, Hermes 3 is built upon the LLaMA 3.1 model. It excels in creativity and storytelling, providing unfiltered outputs without alignment layers. This model is ideal for role-playing and fiction writing.

    LLaMA-3.2 Dark Champion
    Developed by DavidAU, this model utilizes the LLaMA 3.2 Mixture of Experts (MOE) architecture with 18.4 billion parameters. It features a 128,000-token context window, allowing for extensive long-form content processing. High-end GPUs are necessary to run this model efficiently.

    Dolphin 2.9.3 Mistral Nemo 12B
    This 12-billion-parameter model is optimized for local deployment on modern laptops. It offers a 128K context window and excels in coding, instruction following, and general-purpose conversation. Its balance between accessibility and capability makes it a versatile AI assistant.

    Layla
    Layla is a local and uncensored AI entity designed for personal use. It can be run on your local machine, providing unrestricted interactions. The project is available on GitHub, where you can find installation instructions and further details.

    Please ensure that your hardware meets the requirements for these models, as some demand significant computational resources. Additionally, while these models are uncensored, it's important to use them responsibly and be aware of the ethical considerations associated with their deployment.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.