Skip to content
Language models

TwIL 8B ARA Abliterated

an uncensored reasoning model

2 min readupdated

This is a modified version of webAI-Official/TwIL-8B (the verbose variant with long chains of thought) with refusal behavior removed using ARA (Arbitrary-Rank Ablation). Built on IBM Granite 4.2 8B (8.79 billion parameters), it features a 131,072-token context window and generates a reasoning block by default before answering. The author applied ablation across all 40 layers, lowering refusal rates while keeping core logic and coding skills intact. It is aimed at developers and researchers who need a capable local model for complex tasks without strict censorship guardrails.

Key features

  • IBM Granite 4.2 base: a 40-layer dense decoder with a 131K-token context window and a hidden size of 4096.
  • ARA method: full-stack layer ablation with a low Kullback–Leibler divergence (KL divergence 0.0412) to preserve reasoning and baseline capabilities.
  • Lower refusal rate: measured refusals drop to 14%, though the author recommends testing the model on your own specific tasks.
  • GGUF availability: ready-to-use quantizations are available for standard runtimes like Ollama or LM Studio.

What you can do

  • Run locally: load standard GGUF quantizations in LM Studio, Ollama, or KoboldCPP.
  • Solve complex problems: use the built-in reasoning block for math, logic, and coding.
  • Test model behavior: evaluate edge cases where standard base assistants trigger excessive refusals.

Example

The model uses standard chat syntax with a reasoning block before the final response:


User asks a complex question. I need to break down the steps...

Here is the detailed response to your query.
Read moreDetails
Parameters: 8.79B parameters, 40 layers.
Parent model benchmarks: MATH-500 98.2, LiveCodeBench 81.6.
Limitations: released under the webAI Non-Commercial License ver. 1.0.
Notes: KL divergence is 0.0412, refusal rate is 14/100.
Discuss Telegram 0 524

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 16 GB · ready files18 GB
  • FP8 / INT8recommendedweights 8.2 GB10 GB
  • 4-bitweights 4.6 GB6.5 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.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2080 Ti · BF16 — from RTX 4090

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Language models: with a 8192-token context.

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

Run it on Massed Compute

A ready script for the machine we advise above: paste it and the model downloads and starts.

  1. 1. Rent A30

    $0.26 an hour with the coupon. Pick the Ubuntu image when you create the machine.

    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.

    # Austriani/TwIL-8B-ARA-Abliterated — A30, FP8 / INT8
    # NVIDIA driver: CUDA 13 needs 580+. If older or missing, the current nvidia-open is installed (keep yours: KEEP_DRIVER=1)
    DRV=$(nvidia-smi --query-gpu=driver_version --format=csv,noheader 2>/dev/null | head -1 | cut -d. -f1)
    case "$DRV" in ""|*[!0-9]*) DRV=0 ;; esac
    if [ "$DRV" -lt 580 ] && [ -z "${KEEP_DRIVER:-}" ]; then
      . /etc/os-release && \
      curl -fsSLO "https://developer.download.nvidia.com/compute/cuda/repos/ubuntu${VERSION_ID//./}/x86_64/cuda-keyring_1.1-1_all.deb" && \
      sudo dpkg -i cuda-keyring_1.1-1_all.deb && rm -f cuda-keyring_1.1-1_all.deb && \
      sudo apt-get update && \
      sudo DEBIAN_FRONTEND=noninteractive apt-get install -y nvidia-open || \
      { echo "Could not install the NVIDIA driver: see the error above. To keep the current one, run with KEEP_DRIVER=1."; exit 1; }
      echo "NVIDIA driver updated. Reboot the machine (sudo reboot) and run this same script again."
      exit 0
    fi
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv venv ~/vllm --python 3.12 && source ~/vllm/bin/activate
    uv pip install vllm
    vllm serve Austriani/TwIL-8B-ARA-Abliterated --max-model-len 8192 --quantization fp8
  3. 3. Use it

    The server listens on port 8000 and speaks the OpenAI API: http://localhost:8000/v1 on the machine. From your computer: ssh -L 8000:localhost:8000 user@machine-IP.

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.

  • MTP (multi-token prediction)0.34.0from November 16, 2026new release

    1.5–2× generation speed

    If the model has MTP layers (DeepSeek-V3/R1, Qwen3-Next, GLM, MiMo and others), it predicts several tokens per step itself.

    How to enable: vLLM / SGLang: --speculative-config '{"method": "mtp", "num_speculative_tokens": 1}'. Check the model card.

    Earlier generations (2)
    • EAGLE-3 — Speculative decoding with a small draft head. For models without their own MTP layers.
    • EAGLE-2 / Medusa — Earlier draft-head generations. Still work if that is all you find.
  • FP8 KV cache0.34.0from November 16, 2026new release

    2× more context in the same memory

    Keeps the attention cache in 8 bits, so long contexts fit where they did not.

    How to enable: vLLM: --kv-cache-dtype fp8; llama.cpp: -ctk q8_0 -ctv q8_0.

  • vLLM / SGLang0.34.0from November 16, 2026new release

    many times more tokens/s on a server

    Server engines: batching, FlashAttention, FP8 and speculative decoding out of the box.

    How to enable: vllm serve <model> — an OpenAI-compatible API.

    Earlier generations (1)
    • llama.cpp — GGUF models on any GPU, Mac or even CPU; some layers can stay in RAM.

Model family

  1. granite-4.1-8b-basefamily root
  2. granite-4.2-8b
  3. TwIL-8B
  4. TwIL-8B-ARA-Abliterated

New posts about versions of granite-4.1-8b-base (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 3All on Hugging Face
Fine-tunes 6All on Hugging Face

About the model

Austriani/TwIL-8B-ARA-Abliterated
Author
Austriani
Category
Language models
Released
Posted

On NeyroLearn

Hugging Face

5240

Ask AI about this model

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

Comments