{
  "base_model": "Qwen/Qwen3-Reranker-0.6B",
  "format": "coreml-compiled",
  "model": "Qwen3Reranker.mlmodelc",
  "compute_units": "cpuAndNeuralEngine",
  "precision": {
    "weights": "int8 (linear symmetric, per output channel)",
    "activations": "fp16"
  },
  "minimum_os": {
    "macOS": "15.0",
    "iOS": "18.0"
  },
  "tokenizer": "tokenizer.json",
  "add_special_tokens": false,
  "functions": {
    "L128": {
      "tokens": 128
    },
    "L160": {
      "tokens": 160
    },
    "L192": {
      "tokens": 192
    },
    "L224": {
      "tokens": 224
    },
    "L256": {
      "tokens": 256
    },
    "L320": {
      "tokens": 320
    },
    "L384": {
      "tokens": 384
    },
    "L512": {
      "tokens": 512
    },
    "L768": {
      "tokens": 768
    }
  },
  "inputs": {
    "input_ids": {
      "dtype": "int32",
      "shape": [
        1,
        "L"
      ],
      "padding": "left",
      "pad_token_id": 151643
    },
    "mask": {
      "dtype": "float16",
      "shape": [
        1,
        "L"
      ],
      "values": "0 for padding, 1 for tokens"
    }
  },
  "output": {
    "logits": {
      "dtype": "float32",
      "shape": [
        1,
        2
      ],
      "order": [
        "no",
        "yes"
      ],
      "score": "sigmoid(yes - no) = probability that the document is relevant"
    }
  },
  "token_ids": {
    "yes": 9693,
    "no": 2152,
    "pad": 151643
  },
  "prompt": {
    "prefix": "<|im_start|>system\nJudge whether the Document meets the requirements based on the Query and the Instruct provided. Note that the answer can only be \"yes\" or \"no\".<|im_end|>\n<|im_start|>user\n",
    "body": "<Instruct>: {instruction}\n<Query>: {query}\n<Document>: {document}",
    "suffix": "<|im_end|>\n<|im_start|>assistant\n<think>\n\n</think>\n\n",
    "default_instruction": "Given a web search query, retrieve relevant passages that answer the query",
    "prefix_tokens": 39,
    "suffix_tokens": 9,
    "truncate": "truncate the body tokens so prefix + body + suffix <= 768"
  }
}