Skip to content
Other

Arete-27B-Uncensored-V2

чат без отказов и с decision-надстройкой

1 min readupdated

Нецензурированная версия NousResearch/Arete-27B на ~26,9 млрд параметров от SargeDev. V2 добавляет к abliteration из V1 merge Solar decision-LoRA: снятие отказов и обученная логика решений в одной модели.

Ключевые особенности

  • Наследует aggressive abliteration OBLITERATUS из V1 — 15/15 на батарее жёстких compliance-промптов.
  • Solar-JEV decision-LoRA (r32/alpha-64, чекпоинт 1946) влит напрямую по дельтам в 401/401 модуль.
  • MTP/nextn drafter-тензоры возвращены: speculative decoding работает.
  • На 803 промптах Propaganda_Eval: fact score 2,61/5, propaganda 1,1%, ноль жёсткой пропаганды.
  • Decision gate: top-1 0,725 против 0,384 у бэкбона V1.
  • Есть bf16 safetensors (~54 ГБ, 28 шардов) и GGUF-кванты в отдельном репозитории.

Пример

Сценарий из карточки: вопросы об истории цензуры и государственных нарративах. Модель должна не повторять официальные формулировки, а выдавать фактическую картину — на Propaganda_Eval это проверяли отдельно.

ПодробнееКачество после abliteration: coherence 1.0, capability suite intact, perplexity 3,22 → 3,93.
Decision gate уступает huihui-base V3 (0,853) и планке r32 (0,869); автор называет следующий шаг — обучение на самом Arete.
Модель отдаётся as-is: нужна для чат-характера Arete вместе с decision-каркасом.
Формат: bf16 safetensors, 28 шардов, ~54 ГБ; GGUF — отдельно.
Discuss Telegram 1 4

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 51 GB · ready files56 GB
  • FP8 / INT8recommendedweights 25 GB28 GB
  • 4-bitweights 14 GB16 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.

    # SargeDev/Arete-27B-Uncensored-V2
    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 SargeDev/Arete-27B-Uncensored-V2 --local-dir ~/models/Arete-27B-Uncensored-V2
    # 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. Arete-27B
  3. Arete-27B-Uncensored-V2

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

SargeDev/Arete-27B-Uncensored-V2
Author
SargeDev
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

41

Ask AI about this model

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

Comments

Language models

Arete-27B

Arete-27B