{
  "schema_version": 1,
  "published_repository": "TiGa-RCE/Qwen3-Embedding-8B-MLX-Q4",
  "family": "Qwen3-Embedding-8B",
  "variant": "Q4",
  "upstream_repository": "Qwen/Qwen3-Embedding-8B",
  "upstream_revision": "1d8ad4ca9b3dd8059ad90a75d4983776a23d44af",
  "upstream_revision_evidence": "exact source snapshot retained in local Hugging Face download metadata",
  "direct_parent": "TiGa-RCE/Qwen3-Embedding-8B-MLX-BF16",
  "direct_parent_weight_hashes": [
    {
      "file": "model-00001-of-00004.safetensors",
      "sha256": "0a22d58cb671a9049da3505b30b3ba8a200401c4309df397961f61c9c6ab1643",
      "bytes": 4900037603
    },
    {
      "file": "model-00002-of-00004.safetensors",
      "sha256": "18c40e1305906ad2ca3c503efa130f02b20db5940129222cb1728f4a594204ad",
      "bytes": 4915960339
    },
    {
      "file": "model-00003-of-00004.safetensors",
      "sha256": "ce16f97c40ed67ec1b540dcb99259e91a930b431051e50a76350888a2331498c",
      "bytes": 4983068481
    },
    {
      "file": "model-00004-of-00004.safetensors",
      "sha256": "1ac381756541d3601847be3e5ce832a81f308cf8b7afccc1bb65e22c63120cfe",
      "bytes": 335570419
    }
  ],
  "conversion": {
    "method": "Q uniform affine quantization",
    "nominal_bits": 4,
    "group_size": 64,
    "importance_matrix": false,
    "importance_matrix_samples": null,
    "importance_matrix_sequence_length": null,
    "stack": {
      "omlx": "0.5.3",
      "mlx_lm": "0.31.3",
      "mlx": "0.32.0"
    },
    "lossy_parent": false
  },
  "weight_files": [
    {
      "file": "model.safetensors",
      "sha256": "4447f8075ccd210e3c5581722cb2b3b886b12aa6a8ad3b2aaf97eab00b3e6a2c",
      "bytes": 4257149761
    }
  ],
  "evaluation": {
    "pair_count": 24,
    "top1": 1.0,
    "recall_at_5": 1.0,
    "mrr": 1.0,
    "mean_aligned_embedding_cosine_vs_bf16": 0.9731449484825134,
    "minimum_aligned_embedding_cosine_vs_bf16": 0.9609217047691345,
    "score_rmse_vs_bf16": 0.01705232262611389,
    "queries_with_rank_change": 0,
    "gate_passed": false,
    "gate_criteria": {
      "top1_delta_min": 0.0,
      "recall_at_5_delta_min": 0.0,
      "mrr_delta_min": -0.01,
      "minimum_aligned_embedding_cosine_min": 0.99,
      "queries_with_rank_change_max": 2
    }
  },
  "collection": "https://huggingface.co/collections/TiGa-RCE/mlx-embedding-quantization-matrix-q-oq-oqe-at-4-6-8-bit-6a68d11afb238d4fe967d70b"
}