Skip to content
Other

MiniMax-H3-Prompt-Rewriter-Qwen3.8-27B-FL2VA-heretic GGUF - Quantized Weights for Local Run

2 min read

This repository provides ready-to-use static quantizations (compressed weights) of the MiniMax-H3-Prompt-Rewriter-Qwen3.8-27B-FL2VA-heretic model in GGUF format (a format for running large language models efficiently on consumer hardware). Based on the 27-billion-parameter Qwen architecture, the model is optimized for rewriting and enhancing prompts. It supports multimodal workflows thanks to dedicated mmproj files, making it a great fit for developers and enthusiasts looking to run large language models locally using utilities like llama.cpp.

Key features

  • Wide selection of quantizations: ranging from compact Q2_K and Q3_K to high-quality Q6_K and Q8_0, plus the IQ4_XS format.
  • Multimodal support: includes mmproj files (Q8_0 and f16 versions) for working with images.
  • Speed and quality optimization: recommended fast options like Q4_K_M and Q4_K_S are available.
  • GGUF format: full compatibility with popular local clients and inference frameworks.

What you can do

  • Run the model locally on your PC or server using llama.cpp or compatible software, matching the quantization to your VRAM and RAM capacity.
  • Use the model to automatically improve, detail, and rewrite text prompts.
  • Leverage multimodal capabilities by connecting the included mmproj components.
  • Set up prompt generation pipelines with different weight compression levels to save resources.

Example

To run GGUF models and work with multimodal components, use tools like llama.cpp by specifying the path to your chosen quantization weight file and the corresponding mmproj file for image processing.

Read moreDetails
Base model: LeViAtHaN2K/MiniMax-H3-Prompt-Rewriter-Qwen3.8-27B-FL2VA-heretic
Available quantizations: Q2_K (11.2 GB), Q4_K_M (16.8 GB), Q6_K (22.2 GB), Q8_0 (28.7 GB), and others
Multimodal additions: mmproj-Q8_0 (0.7 GB) and mmproj-f16 (1.0 GB)
Quantization author: mradermacher
Discuss Telegram 0

Will it run?

Runs in GGUF Q6_K

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

  • GGUF Q8_0recommendedweights 27 GB · ready files30 GB
  • GGUF Q6_Kweights 21 GB · ready files23 GB
  • GGUF Q5_K_Mweights 18 GB · ready files20 GB
  • GGUF Q5_K_Sweights 17 GB · ready files20 GB
  • GGUF Q4_K_Mweights 16 GB · ready files18 GB
  • GGUF Q4_K_Sweights 15 GB · ready files16 GB
  • GGUF IQ4_XSweights 14 GB · ready files16 GB
  • GGUF Q3_K_Lweights 14 GB · ready files16 GB
  • GGUF Q3_K_Mweights 13 GB · ready files14 GB
  • GGUF Q3_K_Sweights 11 GB · ready files13 GB
  • GGUF Q2_Kweights 10 GB · ready files12 GB

Tip: go 8-bit.GGUF Q8_0 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: GGUF Q8_0 — from RTX 5090

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.

    # mradermacher/MiniMax-H3-Prompt-Rewriter-Qwen3.8-27B-FL2VA-heretic-GGUF
    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 mradermacher/MiniMax-H3-Prompt-Rewriter-Qwen3.8-27B-FL2VA-heretic-GGUF --local-dir ~/models/MiniMax-H3-Prompt-Rewriter-Qwen3.8-27B-FL2VA-heretic-GGUF
    # 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.

MiniMax H3 prompts

All prompts

Ready prompts with the resulting video: copy and try.

Model family

  1. MiniMax-H3-Prompt-Rewriter-Qwen3.8-27B-FL2VAfamily root
  2. MiniMax-H3-Prompt-Rewriter-Qwen3.8-27B-FL2VA-heretic
  3. MiniMax-H3-Prompt-Rewriter-Qwen3.8-27B-FL2VA-heretic-GGUF

New posts about versions of MiniMax-H3-Prompt-Rewriter-Qwen3.8-27B-FL2VA (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 1All on Hugging Face
Fine-tunes 2All on Hugging Face

About the model

mradermacher/MiniMax-H3-Prompt-Rewriter-Qwen3.8-27B-FL2VA-heretic-GGUF
Author
mradermacher
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

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

Comments