Skip to content
Other

MiniMax-H3-ClipProj-GGUF

projection weights for text encoders

2 min readupdated

This repository provides a set of GGUF projection weights (weights quantization and storage format) that enable using more compact Qwen3-VL text encoders (such as the 4B and 8B versions) instead of the standard heavy Qwen3-VL-32B encoder in the MiniMax-H3 model family. Since MiniMax-H3 video, MiniMax-H3-Image, and P-H3-Image models share a 5,120-dimensional text conditioning, a single projection matrix works for all three variants. The generation models themselves, base encoders, and VAE are not included. This tool will be useful for developers and enthusiasts looking to optimize running MiniMax-H3 generative models on available hardware.

Key features

  • Compatibility with MiniMax-H3 video, MiniMax-H3-Image, and P-H3-Image models thanks to a unified text conditioning format.
  • Support for encoders of various sizes: from current versions of Qwen3-VL-4B and Qwen3-VL-8B to legacy Qwen3-VL-2B variants.
  • Availability of two architectural options for matrices: standard projections (such as Ridge) and larger MLP versions.
  • Precise preservation of source weights tensor accuracy without additional quantization.

What you can do

  • Run video or image generation with MiniMax-H3 models using a reduced text encoder to save resources.
  • Build the Romara Labs stable-diffusion.cpp fork with support for custom weights connection patches.
  • Configure command line parameters with --llm and --cliproj flags to integrate matrices into the generation process.

Example

Example of using weights and a text encoder in the command line when starting generation:

--llm /path/to/qwen3-vl-4b.gguf \
--cliproj /path/to/mmh3-4b-ClipProj-v3.1.gguf
Read moreDetails
Projection weights are calibrated on video prompts.
Text input also works for image generation, but quality in this mode has not been systematically tested.
Licensing is subject to the MiniMax H3 Community License Agreement terms.
Discuss Telegram 0 177

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.

    # romara-labs/MiniMax-H3-ClipProj-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 romara-labs/MiniMax-H3-ClipProj-GGUF --local-dir ~/models/MiniMax-H3-ClipProj-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-H3family root
  2. MiniMax-H3
  3. ClipProj-MiniMax-H3
  4. MiniMax-H3-ClipProj-GGUF

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

romara-labs/MiniMax-H3-ClipProj-GGUF
Author
romara-labs
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

1770

Ask AI about this model

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

Comments