{
  "id": "clip-vit-l-14-laion2b-s32b-b82k",
  "identity": {
    "family": "openclip",
    "display_name": "OpenCLIP ViT-L/14 LAION-2B s32B b82K",
    "hub_id": "laion/CLIP-ViT-L-14-laion2B-s32B-b82K",
    "api_name": null,
    "revision": null,
    "variants": [],
    "license": "MIT",
    "commercial_use": true,
    "commercial_use_note": "MIT on the Hugging Face repo. The model card calls the release research-oriented and LAION-2B is uncurated web data; no license restriction on use.",
    "release_date": "2022-09",
    "paper_url": "https://arxiv.org/abs/2212.07143",
    "repo_url": "https://github.com/mlfoundations/open_clip"
  },
  "task": {
    "category": "embedding",
    "modalities_in": [
      "image"
    ],
    "modalities_out": [
      "vector"
    ],
    "best_at": "Image-to-image similarity from a large web-trained ViT-L/14 image tower.",
    "worst_at": "Text-to-image search: the Serve app loads only CLIPVisionModel and returns the raw CLS hidden state, which is not in CLIP's shared text-image space. No extractor writes this vector."
  },
  "io": {
    "input": {
      "image": {
        "resolution": "224x224",
        "color": "RGB",
        "normalize": "CLIP mean/std (CLIPImageProcessor)"
      }
    },
    "output": {
      "vector": {
        "dims": 768,
        "distance": "cosine",
        "type": "dense"
      }
    }
  },
  "mixpeek_wiring": {
    "extractors": [],
    "feature_uris": [],
    "vector_indexes": [],
    "retriever_stages": [],
    "composes_with": [
      "Realtime image embedding through POST /v1/inference only; laion_clip_vit_l_14_v1 is a named inference in the vector models but no extractor declares an index for it"
    ]
  },
  "run": {
    "status": "served",
    "status_reason": "Ray Serve app in the inference registry, image modality only; no extractor uses it.",
    "serve_apps": [
      "laion__CLIP_ViT_L_14_laion2B_s32B_b82K"
    ],
    "runtime": "transformers CLIPVisionModel + CLIPImageProcessor",
    "weights_format": "safetensors",
    "trust_remote_code": false,
    "size_on_disk_mb": 1711,
    "params_millions": 428,
    "preprocessing": "CLIPImageProcessor (resize, center crop 224, CLIP normalization).",
    "postprocessing": "last_hidden_state[:, 0] (CLS before post_layernorm and before the visual projection), L2-normalized. The checkpoint config gives vision hidden_size 1024 and projection_dim 768, so this vector is 1024-d while the registry declares 768.",
    "serving_shape": "both",
    "batch_size": null,
    "serve_batch": null,
    "scale_to_zero": {
      "min_replicas": 0,
      "max_replicas": 1,
      "idle_timeout_s": 600,
      "memory_gb": null,
      "num_cpus": 0.5,
      "num_gpus": {
        "expr": "hardware_config.gpu_fraction if hardware_config.gpu_concurrency > 0 else 0"
      },
      "cold_start_s": {
        "warm_node_s": null,
        "new_node_s": null,
        "source": null
      },
      "cpu_fallback": true
    }
  },
  "cost": {
    "unmeasured": true,
    "measured": null,
    "note": "No committed throughput or cost measurement for this model."
  },
  "evidence": {
    "unmeasured": true,
    "items": []
  },
  "refs": [
    "hf://laion/CLIP-ViT-L-14-laion2B-s32B-b82K@1627032197142fbe2a7cfec626f4ced3ae60d07a"
  ],
  "formats": [
    {
      "ref": "hf://laion/CLIP-ViT-L-14-laion2B-s32B-b82K@1627032197142fbe2a7cfec626f4ced3ae60d07a",
      "format": "safetensors",
      "file": "2 files, e.g. model.safetensors",
      "loader": "transformers / open_clip from_pretrained; Mixpeek Hub import (api/models/huggingface.py)",
      "supported_today": true
    },
    {
      "ref": "hf://laion/CLIP-ViT-L-14-laion2B-s32B-b82K@1627032197142fbe2a7cfec626f4ced3ae60d07a",
      "format": "pytorch_bin",
      "file": "2 files, e.g. open_clip_pytorch_model.bin",
      "loader": "torch.load (pickle); Mixpeek refuses pickle weights (shared/security/torch_pickle.py)",
      "supported_today": false
    }
  ],
  "use_cases": [
    "visual-search",
    "product-search"
  ],
  "recommended": {
    "value": false,
    "for": [],
    "reason": null
  },
  "demos": [],
  "seo": {
    "title": "OpenCLIP ViT-L/14 LAION-2B s32B b82K: embedding model",
    "description": "OpenCLIP ViT-L/14 LAION-2B s32B b82K is a image embedding model (MIT). Served on Mixpeek, scaling to zero when idle.",
    "keywords": [
      "OpenCLIP ViT-L/14 LAION-2B s32B b82K",
      "openclip",
      "embedding",
      "laion/CLIP-ViT-L-14-laion2B-s32B-b82K",
      "visual-search",
      "product-search"
    ],
    "faqs": [
      {
        "q": "What is OpenCLIP ViT-L/14 LAION-2B s32B b82K best at?",
        "a": "Image-to-image similarity from a large web-trained ViT-L/14 image tower."
      },
      {
        "q": "What are the limits of OpenCLIP ViT-L/14 LAION-2B s32B b82K?",
        "a": "Text-to-image search: the Serve app loads only CLIPVisionModel and returns the raw CLS hidden state, which is not in CLIP's shared text-image space. No extractor writes this vector."
      },
      {
        "q": "Can OpenCLIP ViT-L/14 LAION-2B s32B b82K be used commercially?",
        "a": "Yes. MIT on the Hugging Face repo. The model card calls the release research-oriented and LAION-2B is uncurated web data; no license restriction on use."
      },
      {
        "q": "How does OpenCLIP ViT-L/14 LAION-2B s32B b82K run on Mixpeek?",
        "a": "Ray Serve app in the inference registry, image modality only; no extractor uses it."
      }
    ]
  },
  "last_verified": "2026-11-05",
  "sources": [
    "https://huggingface.co/laion/CLIP-ViT-L-14-laion2B-s32B-b82K",
    "https://huggingface.co/laion/CLIP-ViT-L-14-laion2B-s32B-b82K/resolve/main/config.json",
    "https://huggingface.co/api/models/laion/CLIP-ViT-L-14-laion2B-s32B-b82K",
    "https://arxiv.org/abs/2212.07143",
    "server/engine/inference/laion/clip_vit_l_14/services.py",
    "server/engine/inference/laion/clip_vit_l_14/routes.py",
    "server/shared/collection/features/vectors/models.py"
  ],
  "registry": {
    "laion__CLIP_ViT_L_14_laion2B_s32B_b82K": {
      "deployment": "laion__CLIP_ViT_L_14_laion2B_s32B_b82K",
      "service_id": "laion/CLIP-ViT-L-14-laion2B-s32B-b82K",
      "routes": "engine/inference/laion/clip_vit_l_14/routes.py",
      "route_prefix": "/laion_clip_vit_l_14/v1",
      "enabled": false,
      "description": "LAION CLIP ViT-L-14 vision transformer for image embeddings. Generates 768-dimensional embeddings optimized for educational content with better text-in-image understanding (slides, code screenshots).",
      "capabilities": {
        "category": "embedding",
        "modality_input": [
          "image"
        ],
        "modality_output": "vector",
        "supports_batch": true,
        "reranking_mode": null
      },
      "vector_config": {
        "dimensions": 768,
        "distance": "cosine",
        "vector_type": "dense"
      },
      "serve": {
        "min_replicas": 0,
        "max_replicas": 1,
        "idle_timeout_s": 600,
        "target_ongoing_requests": 5,
        "num_cpus": 0.5,
        "num_gpus": {
          "expr": "hardware_config.gpu_fraction if hardware_config.gpu_concurrency > 0 else 0"
        },
        "memory_gb": null,
        "pinnable": false,
        "placement": "worker",
        "enabled": true
      },
      "interactive_twin": false,
      "interactive_serve": null,
      "schemas": {
        "input": "shared.inference.laion.clip_vit_l_14.models.InferenceRequest",
        "output": "shared.inference.laion.clip_vit_l_14.models.InferenceResponse",
        "config": "shared.inference.laion.clip_vit_l_14.models.InferenceConfigs"
      }
    }
  },
  "templates": [],
  "usage": {
    "namespaces": null,
    "collections": null,
    "uncounted": true,
    "templates": 0,
    "counted_at": null,
    "source": null
  },
  "facets": {
    "status": "served",
    "category": "embedding",
    "family": "openclip",
    "license": "MIT",
    "commercial_use": true,
    "modalities_in": [
      "image"
    ],
    "modalities_out": [
      "vector"
    ],
    "use_cases": [
      "visual-search",
      "product-search"
    ],
    "recommended": false,
    "release_date": "2022-09",
    "params_millions": 428,
    "size_on_disk_mb": 1711,
    "vector_dims": 768,
    "vector_distance": "cosine",
    "weights_format": "safetensors",
    "formats": [
      "pytorch_bin",
      "safetensors"
    ],
    "loadable_today": true,
    "trust_remote_code": false,
    "cpu_fallback": true,
    "num_gpus": null,
    "memory_gb": null,
    "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": 3
  },
  "rank_score": 3
}