Skip to content
Other

Qwen3.8-27B AEON ULTIMATE UNCENSORED

смешанная квантизация для MLX

1 min readupdated

Это квантованная версия модели Qwen3.8-27B AEON ULTIMATE UNCENSORED, подготовленная через oQ (oMLX v0.7.4.dev1). Чекпоинт использует mixed-precision квантизацию: средняя эффективная точность — 3.678 бита на вес при 27.781B логических весов.

Ключевые особенности

  • Смешанная точность: слои хранятся в форматах от affine 2-bit до BF16.
  • Основная доля весов — affine 3-bit, group 128: 73.796%.
  • Общий размер чекпоинта — 11.894 GiB.
  • В составе есть vision-компонент на 0.858 GiB и MTP head на 0.293 GiB.
  • Квантование выполнено инструментом oQ из oMLX v0.7.4.dev1.

Пример

Сценарий: загрузите чекпоинт в MLX-совместимый рантайм и попросите модель разобрать распределение битов. Например: «Объясни, какие компоненты этого чекпоинта занимают больше всего места и почему средняя точность составляет 3.678 бита на вес».

ПодробнееСредняя эффективная точность — 3.678 бит/вес, логических весов — 27.781B.
Хранилище: 11.894 GiB; language model — 10.743 GiB, vision — 0.858 GiB, MTP head — 0.293 GiB.
Бенчмарков и требований к VRAM в карточке нет — только данные квантизации.
Discuss Telegram 0 26

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 12 GB · ready files14 GB
  • FP8 / INT8recommendedweights 5.9 GB7.0 GB
  • 4-bitweights 3.3 GB4.2 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 5060 Ti 8GB · BF16 — from RTX 5080

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

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.

    # chogeha/Qwen3.8-27B-AEON-ULTIMATE-UNCENSORED-oQ2.5e-mtp
    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 chogeha/Qwen3.8-27B-AEON-ULTIMATE-UNCENSORED-oQ2.5e-mtp --local-dir ~/models/Qwen3.8-27B-AEON-ULTIMATE-UNCENSORED-oQ2.5e-mtp
    # 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

New posts about versions of Qwen3.8-27B-AEON-ULTIMATE-UNCENSORED-oQ2.5e-mtp (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

chogeha/Qwen3.8-27B-AEON-ULTIMATE-UNCENSORED-oQ2.5e-mtp
Author
chogeha
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

260

Ask AI about this model

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

Comments