Skip to content
Other

MiniMax-H3 FP8

квантованная сборка для diffusers

1 min read

Репозиторий addlabsviral/minimax-h3-fp8 — FP8 weight-only квантование MiniMax-H3 через torchao Float8WeightOnlyConfig (e4m3). В трансформере FL2VA / i2v веса хранятся в FP8, вычисления идут в bf16, а turbo LoRA уже вшита. Модель на 33.12B параметров; VAE, scheduler, tokenizer и processor оставлены без изменений в bf16.

Ключевые особенности

  • FP8 weight-only для transformer/: меньше вес, compute bf16.
  • Turbo LoRA уже свёрнута в веса — отдельно подключать не нужно.
  • Настройки VAE, scheduler, tokenizer и processor не квантованы.
  • В этот же репозиторий обещают добавить text_encoder/ (Qwen3-VL) и второй transformer.
  • Загрузка через diffusers ModularPipeline с workflow="fl2va".

Пример

import torch
from diffusers import ModularPipeline
pipe = ModularPipeline.from_pretrained("addlabsviral/minimax-h3-fp8", workflow="fl2va")
# transformer loads FP8 automatically via config.json (torchao)
ПодробнееПараметров: 33.12B.
Квантован только transformer; text_encoder и второй transformer пока в планах.
VAE, scheduler, tokenizer и processor — bf16 без изменений.
Апстрим: MiniMaxAI/MiniMax-H3.
Discuss Telegram 0 190

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.

  • F8_E4M3recommendedweights 32 GB · ready files35 GB
  • 4-bitweights 17 GB20 GB

Tip: go 8-bit.F8_E4M3 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: Won't fit any home GPU — renting is cheaper.

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.

    # addlabsviral/minimax-h3-fp8
    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 addlabsviral/minimax-h3-fp8 --local-dir ~/models/minimax-h3-fp8
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to minimax-h3-fp8
  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.

MiniMax H3 prompts

All prompts

Ready prompts with the resulting video: copy and try.

Model family

New posts about versions of minimax-h3-fp8 (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

addlabsviral/minimax-h3-fp8
Author
addlabsviral
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

1900

Ask AI about this model

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

Comments