CivArchive
    Clip - CLIP ViT-L/14
    Preview 135264044

    CLIP (Contrastive Language-Image Pre-Training) is a neural network trained on a variety of (image, text) pairs. It can be instructed in natural language to predict the most relevant text snippet, given an image, without directly optimizing for the task, similarly to the zero-shot capabilities of GPT-2 and 3. We found CLIP matches the performance of the original ResNet50 on ImageNet “zero-shot” without using any of the original 1.28M labeled examples, overcoming several major challenges in computer vision.

    https://github.com/openai/CLIP/

    Description

    Comments (1)

    FelldudeJul 23, 2026
    CivitAI

    Of note this is the "pruned" version lacking the vision model and would likely only be used in SD 1.5, SDXL, FLUX or any diffusion DiT workflow that uses "CLIP-L"

    The pruned FP32 version can be found at:

    https://huggingface.co/stable-diffusion-v1-5/stable-diffusion-v1-5/tree/main/text_encoder

    The full model with vision by open-ai can be found at:

    https://huggingface.co/openai/clip-vit-large-patch14/tree/main

    Other
    OpenAI
    by CivitaiOfficial
    Download (Beta) View on CivitAI
    base model

    Details

    Downloads
    17
    Platform
    CivitAI
    Platform Status
    Available
    Created
    7/9/2026
    Updated
    8/18/2026
    Deleted
    -

    Files

    clip_VitL14.safetensors

    Size:
    234.74 MB
    SHA256:
    660c6f5b1abae9dc498ac2d21e1347d2abdb0cf6c0c0c8576cd796491d9a6cdd

    Mirrors

    HuggingFace (1961 mirrors)
    CivitAI (29 mirrors)
    TensorFiles (1 mirrors)
    CivitasBay (1 mirrors)
    ModelScope (6 mirrors)
    ModelScope CN (6 mirrors)