Modelsopenaiclip-vit-large-patch14-336
provideropenai /

clip-vit-large-patch14-336

0.175 DZD/ second

A zero-shot-image-classification model released by OpenAI. The clip-vit-large-patch14-336 model was trained from scratch on an unknown dataset and achieves unspecified results on the evaluation set. The model's intended uses and limitations, as well as its training and evaluation data, are not provided. The training procedure used an unknown optimizer and precision, and the framework versions included Transformers 4.21.3, TensorFlow 2.8.2, and Tokenizers 0.12.1.

Public
OpenAIOpenAI
clip-vit-large-patch14-336
ArchitectureDense
Context Windowimage-classification

📌 Overview

The clip‑vit‑large‑patch14‑336 is a variant of OpenAI’s CLIP (Contrastive Language–Image Pretraining) model.
It aligns images and text in a shared embedding space, enabling tasks like zero‑shot classification, image‑text retrieval, and multimodal reasoning.

FieldDetails
DeveloperOpenAI
Model typeVision–Language Model
ArchitectureVision Transformer (ViT‑Large, patch size 14)
Input resolution336 × 336 pixels
OutputJoint text–image embeddings
Release date2021
LicenseOpen‑source (MIT)

🎯 Intended Use

  • Zero‑shot image classification using natural language prompts.
  • Image–text similarity and retrieval.
  • Multimodal applications combining vision and language.
  • Research in representation learning and cross‑modal tasks.

Out‑of‑scope:

  • Safety‑critical decision making without human oversight.
  • Applications requiring perfect semantic alignment in all contexts.

⚙️ Features

  • Pretrained on 400M image–text pairs.
  • Strong generalization across unseen categories.
  • Works with natural language prompts instead of fixed labels.
  • Supports flexible multimodal reasoning.

📊 Performance

  • Competitive zero‑shot accuracy on ImageNet and other benchmarks.
  • Robust across diverse datasets and domains.
  • Limitations: may reflect biases in training data, and performance depends on prompt quality.

🛡️ Responsibility & Safety

  • Outputs should be reviewed for fairness and bias.
  • Not suitable for sensitive or high‑risk applications without validation.
  • Recommended to pair with human oversight in production systems.