Skip to content
Other

Ostrich-27B-261118-Qwen3.8-Abliterated

a new uncensored language model for private consultations and analytics

2 min readupdated

Built on the efficient Qwen3.8-27B LLM (large language model), Ostrich-27B-261118-Qwen3.8-Abliterated is trained on specialized corpora covering scientific papers, health, nutrition, and technology. Its defining trait is that it is abliterated (built-in refusals and censorship have been completely stripped out), giving you direct, detailed answers on any sensitive or complex topic. It runs efficiently on consumer hardware, making it a strong fit for researchers, homeschooling parents, and anyone looking for a private digital assistant.

Key features

  • Fully uncensored and refusal-free (abliterated) to answer any question without blocks.
  • Specialized training on scientific papers, medical traditions, nutrition, and technology.
  • Uses the cpt6 adapter family (vibe, scientific papers, reports) for deep domain adaptation.
  • Strong scores across alternative medicine, Bitcoin, and nutrition benchmarks.

What you can do

  • Run a private local chatbot to answer questions about health, herbs, and biology.
  • Conduct deep analysis of primary scientific papers and technical reports using the built-in adapters.
  • Use the model as an educational tool for homeschooling and in-depth research.

Example

Ask the model about traditional wellness practices, nutritional principles, or information security technologies to get a direct, expert breakdown free of censorship restrictions.

Read moreDetails
Benchmarks: overall geometric mean is 82% (Faith: 94%, Health: 80%, Nutrition: 81%, Bitcoin: 84%, Alt-Med: 91%).
Hardware requirements: roughly 18 GB VRAM at 4-bit quantization (fits on a single consumer GPU).
Training details: late layers (>=48) frozen to preserve stack stability; scientific paper corpus chunked into 4,160-token documents.
Testing: passed 20 out of 20 test prompts with zero refusals (100% compliance).
Discuss Telegram 1 17

Will it run?

Runs in 4-bit

Full precision does not fit, a quantised version does: quality barely changes down to 8 bits, more noticeably at 4 bits.

  • BF16weights 52 GB · ready files57 GB
  • FP8 / INT8recommendedweights 26 GB29 GB
  • 4-bitweights 15 GB17 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # etemiz/Ostrich-27B-261118-Qwen3.8-Abliterated
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download etemiz/Ostrich-27B-261118-Qwen3.8-Abliterated --local-dir ~/models/Ostrich-27B-261118-Qwen3.8-Abliterated
    # The model is in ~/models. How to run it is in the model card (link above).
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

  1. Qwen3.8-27Bfamily root
  2. Ostrich-27B-260903-Qwen3.8
  3. Ostrich-27B-261008-Qwen3.8-Abliterated
  4. Ostrich-27B-261028-Qwen3.8-Abliterated
  5. Ostrich-27B-261118-Qwen3.8-Abliterated

New posts about versions of Qwen3.8-27B (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 2All on Hugging Face

About the model

etemiz/Ostrich-27B-261118-Qwen3.8-Abliterated
Author
etemiz
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

171

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments