# facebookresearch/MetaCLIP

NeurIPS 2025 Spotlight; ICLR2024 Spotlight; CVPR 2024; EMNLP 2024

Repository: https://github.com/facebookresearch/MetaCLIP
Canonical: https://ross.abutalabs.com/products/metaclip
Language: Python
License: NOASSERTION
License Family: other
Last push: 2026-08-11T20:09:16+00:00

## Health v2 (maintenance only)
Score: 82/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 97, release rhythm 65, longevity 76
- inputs: {"age_days": 1070, "days_push": 22, "days_rel": 22, "gap_med": 262, "n_releases_24m": 2}
- flags: no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1854, forks 80 (observed 2026-08-28T04:05:44.608757+00:00)

## What it is
Meta's research code and models for Meta CLIP, a reimplementation and scaling recipe for CLIP-style contrastive vision-language models, including Meta CLIP 2 for worldwide multilingual data curation and training. It provides pretrained checkpoints, training/eval code, and metadata curation pipelines released alongside NeurIPS/ICLR/CVPR papers.

## Use cases
- train a CLIP model on curated image-text data
- get multilingual CLIP embeddings for images and text
- run zero-shot image classification with a pretrained CLIP model
- curate large-scale non-English training data for vision-language models
- fine-tune or evaluate CLIP models with open_clip or Hugging Face
- reproduce Meta CLIP research results from the papers

## When to choose
- you need state-of-the-art multilingual CLIP embeddings
- you want to reproduce or extend Meta's CLIP data curation research
- you need pretrained vision-language backbones for downstream tasks like retrieval or zero-shot classification

## When to avoid
- you need a production-ready inference service rather than research code
- you lack GPU resources for large-scale training
- you need a lightweight plug-and-play image classifier without vision-language modeling

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, image-processing, nlp, data-science
- domain: machine-learning, deep-learning, computer-vision, artificial-intelligence
- platform: python, cross-platform
- tags: clip, contrastive-learning, multilingual, vision-language-models, data-curation, image-text-retrieval, zero-shot-classification, research-code, facebook-research, natural-language-processing, gpu, linux

## Member repositories
- facebookresearch/MetaCLIP (main) score 82

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:44.608757+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:16:48.477277+00:00, confidence not recorded.
  - readme: https://github.com/facebookresearch/MetaCLIP (fetched 2026-08-28T04:05:44.608757+00:00, sha 1419a31133ab)
- Data as of 2026-08-30T08:39:29.467469+00:00.