CivArchive
    Qwen-Image GGUF 3Q_K_M - 10G VRAM - VAE
    NSFW

    This is a 3-bit quantized GGUF conversion of the Qwen/Qwen-Image model, released by City96 and mirrored here for convenience. The Q3_K_M variant is optimized for GPUs with at least 10GB VRAM,

    • Tested VRAM Usage:

      • Ubuntu, Firefox (8 tabs): ~8.8 - 9.2 GB VRAM (CFG 3, 20 Steps, uni_pc, normal, 5.4s/it)

      • Windows 11, Brave (1 tab), MiniConda, GGUF excluded from Windows Defender: ~9.6 GB VRAM

      Tips: Offload monitor tasks to an integrated GPU to free up VRAM. Runs smoothly on Linux or Windows

    Description

    VAE for Qwen Image

    FAQ

    Checkpoint
    Other
    by civitai
    Download (Beta) View on CivitAI
    qwen
    base model
    text-to-image
    t2i

    Details

    Downloads
    495
    Platform
    CivitAI
    Platform Status
    Available
    Created
    8/8/2025
    Updated
    11/13/2026
    Deleted
    -

    Files

    qwen_image_vae.safetensors

    Size:
    242.05 MB
    SHA256:
    a70580f0213e67967ee9c95f05bb400e8fb08307e017a924bf3441223e023d1f

    Mirrors

    HuggingFace (368 mirrors)
    CivitAI (311 mirrors)
    TensorFiles (1 mirrors)
    ModelScope (3 mirrors)
    ModelScope CN (7 mirrors)