{
  "id": "dfn2b-clip-vit-l-14",
  "identity": {
    "family": "dfn",
    "display_name": "DFN2B CLIP ViT-L/14",
    "hub_id": "apple/DFN2B-CLIP-ViT-L-14",
    "api_name": null,
    "revision": null,
    "variants": [],
    "license": "apple-amlr",
    "commercial_use": false,
    "commercial_use_note": "Apple AMLR (sample code) license: research use; the weights are not licensed for commercial products.",
    "release_date": "2023-10-30",
    "paper_url": "https://arxiv.org/abs/2309.17425",
    "repo_url": "https://github.com/apple/ml-mobileclip"
  },
  "task": {
    "category": "embedding",
    "modalities_in": [
      "image",
      "text"
    ],
    "modalities_out": [
      "vector"
    ],
    "best_at": "Published COCO 5k recall@1 59.6 and Flickr30k 84.7 at ViT-L size.",
    "worst_at": "Non-commercial license; ships no preprocessor_config.json (Mixpeek builds one from open_clip_config.json); 77-token text."
  },
  "io": {
    "input": {
      "image": {
        "resolution": "224x224",
        "color": "RGB",
        "normalize": "CLIP mean (0.4815, 0.4578, 0.4082), std (0.2686, 0.2613, 0.2758)"
      },
      "text": {
        "max_tokens": 77
      },
      "prompts": "captions as written; CLIP BPE tokenizer (lowercases), 77 tokens"
    },
    "output": {
      "vector": {
        "dims": 768,
        "distance": "cosine",
        "type": "dense"
      },
      "note": "image and text vectors share one space; L2-normalize before cosine"
    }
  },
  "mixpeek_wiring": {
    "extractors": [],
    "feature_uris": [],
    "vector_indexes": [],
    "retriever_stages": [
      "feature_search",
      "mmr"
    ],
    "composes_with": [
      "video frames (VideoFrameExpander) -> image vectors -> feature_search with a text query",
      "detector boxes (RT-DETR, D-FINE) -> crops -> image vectors per object",
      "image vectors + E5 text vectors on captions -> feature_search fusion (dbsf)"
    ]
  },
  "run": {
    "status": "plugin",
    "status_reason": "Imported from the Hugging Face Hub as an hf:// extractor and served by the transformers image-text loader, safetensors only, trust_remote_code false. GPU recommended at this size.",
    "serve_apps": [],
    "runtime": "transformers CLIPModel + CLIPProcessor",
    "weights_format": "safetensors",
    "trust_remote_code": false,
    "size_on_disk_mb": 1710,
    "params_millions": 428,
    "preprocessing": "Resize to 224x224 without a crop (open_clip resize_mode squash), CLIP mean/std normalize; tokenize to 77.",
    "postprocessing": "L2-normalize; cosine similarity.",
    "serving_shape": "both",
    "batch_size": 32,
    "serve_batch": null,
    "scale_to_zero": {
      "min_replicas": 0,
      "max_replicas": 1,
      "idle_timeout_s": 600,
      "memory_gb": 6,
      "num_cpus": 2,
      "num_gpus": 1,
      "cold_start_s": {
        "warm_node_s": null,
        "new_node_s": null
      },
      "cpu_fallback": true
    }
  },
  "cost": {
    "unmeasured": true,
    "measured": null,
    "note": "No committed measurement on Mixpeek yet; the SOTA bench measures it."
  },
  "evidence": {
    "unmeasured": true,
    "items": []
  },
  "refs": [
    "hf://apple/DFN2B-CLIP-ViT-L-14@d431cb9b550f020c753196f5871c1b7c0dfae5e8"
  ],
  "formats": [
    {
      "ref": "hf://apple/DFN2B-CLIP-ViT-L-14@d431cb9b550f020c753196f5871c1b7c0dfae5e8",
      "format": "safetensors",
      "file": "2 files, model.safetensors (transformers) and open_clip_pytorch_model.safetensors (open_clip, not read)",
      "loader": "transformers from_pretrained; Mixpeek Hub import",
      "supported_today": true
    }
  ],
  "use_cases": [
    "visual-search"
  ],
  "recommended": {
    "value": false,
    "for": [],
    "reason": null
  },
  "demos": [],
  "seo": {
    "title": "DFN2B CLIP ViT-L/14: embedding model",
    "description": "DFN2B CLIP ViT-L/14 is a image and text embedding model (apple-amlr, non-commercial). Can run on Mixpeek as an imported Hugging Face model.",
    "keywords": [
      "DFN2B CLIP ViT-L/14",
      "dfn",
      "embedding",
      "apple/DFN2B-CLIP-ViT-L-14",
      "visual-search"
    ],
    "faqs": [
      {
        "q": "What is DFN2B CLIP ViT-L/14 best at?",
        "a": "Published COCO 5k recall@1 59.6 and Flickr30k 84.7 at ViT-L size."
      },
      {
        "q": "What are the limits of DFN2B CLIP ViT-L/14?",
        "a": "Non-commercial license; ships no preprocessor_config.json (Mixpeek builds one from open_clip_config.json); 77-token text."
      },
      {
        "q": "Can DFN2B CLIP ViT-L/14 be used commercially?",
        "a": "No. Apple AMLR (sample code) license: research use; the weights are not licensed for commercial products."
      },
      {
        "q": "How does DFN2B CLIP ViT-L/14 run on Mixpeek?",
        "a": "Imported from the Hugging Face Hub as an hf:// extractor and served by the transformers image-text loader, safetensors only, trust_remote_code false. GPU recommended at this size."
      },
      {
        "q": "What does DFN2B CLIP ViT-L/14 output?",
        "a": "A 768-dimensional dense vector, compared by cosine similarity."
      }
    ]
  },
  "last_verified": "2026-11-12",
  "sources": [
    "https://huggingface.co/apple/DFN2B-CLIP-ViT-L-14",
    "https://arxiv.org/abs/2309.17425",
    "https://github.com/apple/ml-mobileclip"
  ],
  "templates": [],
  "usage": {
    "namespaces": null,
    "collections": null,
    "uncounted": true,
    "templates": 0,
    "counted_at": "2026-11-06T13:42:25Z"
  },
  "facets": {
    "status": "plugin",
    "category": "embedding",
    "family": "dfn",
    "license": "apple-amlr",
    "commercial_use": false,
    "modalities_in": [
      "image",
      "text"
    ],
    "modalities_out": [
      "vector"
    ],
    "use_cases": [
      "visual-search"
    ],
    "recommended": false,
    "release_date": "2023-10-30",
    "params_millions": 428,
    "size_on_disk_mb": 1710,
    "vector_dims": 768,
    "vector_distance": "cosine",
    "weights_format": "safetensors",
    "formats": [
      "safetensors"
    ],
    "loadable_today": true,
    "trust_remote_code": false,
    "cpu_fallback": true,
    "num_gpus": 1,
    "memory_gb": 6,
    "min_replicas": 0,
    "serving_shape": "both",
    "served": false,
    "ref_schemes": [
      "hf"
    ],
    "has_evidence": false,
    "cost_measured": false,
    "template_count": 0,
    "usage_namespaces": null,
    "usage_collections": null,
    "usage_uncounted": true,
    "has_demo": false,
    "rank_score": 1
  },
  "rank_score": 1
}
