Skip to content
Images

MiniMax-H3-Image-Pruned

меньше веса, те же пиксели

1 min readupdated

WaveCut опубликовал прунинг-версию MiniMax-H3-Image для text-to-image. Из чекпоинта удалены веса, которые инференс не использует: размер снизился со 135,3 до 94,1 ГБ, а картинки на той же GPU остаются бит-в-бит идентичными оригиналу. Ничего не квантовали и не дообучали.

Ключевые особенности

  • Удалены AdaLN-проекции на 13,0B параметров — их заменили таблицами на 116 МБ.
  • Текстовый энкодер Qwen3-VL-32B урезан с 64 до 50 decoder layers; lm_head тоже убран.
  • Трансформер: 33,1B → 20,1B параметров, 66,3 → 40,3 ГБ.
  • Мёртвые строки последнего блока пропускаются только на вычислениях, не на весах.
  • На RTX PRO 6000 (96 ГБ) ускорение за счёт того, что модели помещаются на GPU без offload.

Пример

import torch
from minimax_h3_image_pruned import load_pipeline

pipe = load_pipeline("WaveCut/MiniMax-H3-Image-Pruned")
image = pipe(prompt="The image is a top-down photograph of ...",
             height=3552, width=4736,
             generator=torch.Generator().manual_seed(42),
             output="images")[0]
ПодробнееНа RTX PRO 6000 (96 ГБ), 4736×3552: 26,3 с против 101,2 с у оригинала; пик памяти — 92,6 GiB.
Работает только на шестишаговой сетке сэмплирования; SGLang и ComfyUI этот формат не загружают.
Таблицы собраны на RTX PRO 6000 Blackwell — на другом типе GPU возможно расхождение в пределах bfloat16.
Лицензия MiniMax-H3-Image Community: разрешённая территория не включает ЕС, Великобританию, Корею и США.
Примеры
Discuss Telegram 1 47

Will it run?

Runs in FP8 / INT8

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

  • BF16weights 37 GB · ready files43 GB
  • FP8 / INT8recommendedweights 19 GB22 GB
  • 4-bitweights 11 GB13 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 4090

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.

    # WaveCut/MiniMax-H3-Image-Pruned
    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 WaveCut/MiniMax-H3-Image-Pruned --local-dir ~/models/MiniMax-H3-Image-Pruned
    # 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.

  • SageAttention 3v2.2.0from October 28, 2025

    even faster on RTX 50 / Blackwell

    FP4 attention for Blackwell. Faster than SageAttention 2, but the authors suggest v2 when accuracy matters.

    How to enable: Build from the sageattention3_blackwell folder of the repo; in ComfyUI via KJNodes.

    Earlier generations (2)
    • SageAttention 2++ — Quantised (INT8/FP8) attention with no visible quality loss. The main speed-up for image and video generation.
    • SageAttention 1 — First version: INT8 for Q·K. Still works, but 2++ is faster and more accurate.
  • 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 (3)
    • 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.
    • xFormers — The older way to speed up attention, works almost everywhere. Sage or Flash are faster where available.
  • cache-dit (DBCache)v1.5.6from November 20, 2026new release

    1.5–3× at nearly the same quality

    Reuses block outputs between diffusion steps. Supports most DiT models.

    How to enable: pip install cache-dit; one line cache_dit.enable_cache(pipe) for diffusers.

    Earlier generations (2)
    • MagCache — Skips steps by magnitude of change. More accurate than TeaCache, no calibration.
    • TeaCache — The first popular diffusion step cache. Works, newer methods are more accurate.
  • Nunchaku (SVDQuant 4-bit)v1.2.1from January 25, 2026

    ≈3× faster and 3.5× less memory

    4-bit versions of popular image models with almost no quality loss (FP4 on RTX 50).

    How to enable: Ready weights for supported models; in ComfyUI use ComfyUI-nunchaku.

  • 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.

MiniMax H3 prompts

All prompts

Ready prompts with the resulting video: copy and try.

Model family

  1. MiniMax-H3family root
  2. MiniMax-H3-Image
  3. MiniMax-H3-Image-Pruned

New posts about versions of MiniMax-H3 (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 12All on Hugging Face
Merges 3All on Hugging Face

About the model

WaveCut/MiniMax-H3-Image-Pruned
Author
WaveCut
Category
Images
Released
Posted

On NeyroLearn

Hugging Face

471

Ask AI about this model

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

Comments