Skip to content
Other

Qwen3.8-Flash-Next-oQ2-mtp

180B в MLX с 2,85 бит/вес

1 min read

Квантованный чекпоинт Qwen3.8-Flash-Next от timesmith для библиотеки MLX. Смешанное квантование выполнено через oQ (oMLX v0.7.3): 180B параметров упакованы примерно в 59,7 GiB.

Ключевые особенности

  • Эффективная точность — 2,790 бит/вес без N-gram embedding и 2,850 бит/вес для полного чекпоинта.
  • Основная масса весов — affine 2-bit group 64: 120,803B параметров, 35,158 GiB и 67,1% от общего объёма.
  • N-gram embedding — 51,2B параметров в affine 2-bit group 32, занимает 17,881 GiB.
  • Общий размер чекпоинта — 64,120 GB / 59,717 GiB при 180B параметров.
  • В составе есть vision-компонент на 0,449B и MTP head на 2,607B.
  • Формат — MLX; квантование oQ смешивает 2, 3, 4, 5, 6, 8-bit и BF16.

Пример

Сценарий из карточки — локальная работа с 180B-чекпоинтом в MLX: языковая часть занимает 42,519 GB, N-gram embedding — 19,2 GB, vision — 0,898 GB, MTP head — 1,503 GB. Итоговый вес — около 59,7 GiB.

ПодробнееЭффективно 2,790 бит/вес без N-gram-таблицы; 2,850 бит/вес для полного чекпоинта.
Общий вес — 64,120 GB (59,717 GiB).
Языковая модель — 125,744B параметров, N-gram embedding — 51,200B, vision — 0,449B, MTP head — 2,607B.
В карточке нет бенчмарков, лицензии и инструкции запуска.
Discuss Telegram 0 70

Will it run?

Working out the memory needs

The bot is still collecting this model's weight sizes; the estimate will appear here in a few minutes. Meanwhile you can rent a GPU with headroom.

Massed ComputeOpen Massed Compute

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.

    # timesmith/Qwen3.8-Flash-Next-oQ2-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 timesmith/Qwen3.8-Flash-Next-oQ2-mtp --local-dir ~/models/Qwen3.8-Flash-Next-oQ2-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-Flash-Next-oQ2-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

timesmith/Qwen3.8-Flash-Next-oQ2-mtp
Author
timesmith
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

700

Ask AI about this model

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

Comments

Computer vision

UAV-PUR-8B

UAV-PUR-8B