{
  "schema_version": 1,
  "family": "confucius4_r2t2",
  "display_name": "Confucius4-R2T2",
  "description": "NetEase Youdao Confucius4-R2T2 real-time ASR: a Qwen3-ASR fine-tune with Longest Stable Prefix (LSP) streaming. Low-latency append-only streaming from 80 ms to 2 s chunks, context and hotword prompts, 30 languages.",
  "category": "asr",
  "status": "community",
  "tasks": [
    "asr"
  ],
  "modes": [
    "offline",
    "streaming"
  ],
  "languages": [
    "zh",
    "en",
    "yue",
    "ar",
    "de",
    "fr",
    "es",
    "pt",
    "id",
    "it",
    "ko",
    "ru",
    "th",
    "vi",
    "ja",
    "tr",
    "hi",
    "ms",
    "nl",
    "sv",
    "da",
    "fi",
    "pl",
    "cs",
    "fil",
    "fa",
    "el",
    "hu",
    "mk",
    "ro"
  ],
  "capabilities": {
    "asr": [
      "partial_results"
    ]
  },
  "runtime": {
    "tags": [
      "gguf",
      "stream"
    ]
  },
  "ui": {
    "recommended_package": "confucius4_r2t2_q8_0",
    "tags": [
      "ASR",
      "Stream"
    ],
    "docs": [
      "docs/models/r2t2.md",
      "docs/asr.md"
    ]
  },
  "package_defaults": {
    "download": {
      "kind": "huggingface_snapshot",
      "repo": "netease-youdao/Confucius4-R2T2",
      "revision": "main",
      "gated": false
    }
  },
  "packages": [
    {
      "id": "confucius4_r2t2_q8_0",
      "display_name": "Confucius4-R2T2 Q8_0 GGUF",
      "default": true,
      "format": "gguf",
      "precision": "q8_0",
      "target_directory": "Confucius4-R2T2-GGUF",
      "files": [
        "r2t2-q8_0.gguf"
      ],
      "download": {
        "kind": "huggingface_snapshot",
        "repo": "davidxifeng/Confucius4-R2T2-gguf"
      }
    },
    {
      "id": "confucius4_r2t2_f16",
      "display_name": "Confucius4-R2T2 F16 GGUF",
      "format": "gguf",
      "precision": "f16",
      "target_directory": "Confucius4-R2T2-GGUF",
      "files": [
        "r2t2-f16.gguf"
      ],
      "download": {
        "kind": "huggingface_snapshot",
        "repo": "davidxifeng/Confucius4-R2T2-gguf"
      }
    },
    {
      "id": "confucius4_r2t2_q4_k_m",
      "display_name": "Confucius4-R2T2 Q4_K_M GGUF",
      "description": "Community Q4_K_M quantization by NairoDorian (Hugging Face: Nairod785), pinned to a fixed revision; quantization report in QUANTIZATION.md. Quantized derivative work of netease-youdao/Confucius4-R2T2 under the NetEase Youdao Model Use License - see LICENSE, LICENSE_zh and NOTICE in the download repo.",
      "format": "gguf",
      "precision": "q4_k_m",
      "target_directory": "Confucius4-R2T2-Q4_K_M-GGUF",
      "files": [
        "r2t2-q4_k_m.gguf"
      ],
      "download": {
        "kind": "huggingface_snapshot",
        "repo": "Nairod785/Confucius4-R2T2-Q4_K_M-GGUF",
        "revision": "b1ea19256fb77a8d8ab7b091dc75378e15952605",
        "gated": false
      }
    },
    {
      "id": "confucius4_r2t2_safetensors",
      "display_name": "Confucius4-R2T2 (HF safetensors)",
      "format": "safetensors",
      "precision": "native",
      "target_directory": "Confucius4-R2T2",
      "files": [
        "added_tokens.json",
        "chat_template.json",
        "config.json",
        "generation_config.json",
        "merges.txt",
        "model.safetensors",
        "preprocessor_config.json",
        "special_tokens_map.json",
        "tokenizer.json",
        "tokenizer_config.json",
        "vocab.json"
      ],
      "download": {
        "kind": "huggingface_snapshot",
        "repo": "netease-youdao/Confucius4-R2T2"
      }
    }
  ],
  "sources": [
    {
      "format": "gguf",
      "roots": {
        "model": ".",
        "weights": "$gguf"
      },
      "files": {
        "config": "model:config.json",
        "generation_config": "model:generation_config.json",
        "tokenizer_config": "model:tokenizer_config.json"
      },
      "optional_files": {
        "preprocessor_config": "model:preprocessor_config.json",
        "processor_config": "model:processor_config.json",
        "chat_template": "model:chat_template.json",
        "chat_template_jinja": "model:chat_template.jinja",
        "vocab": "model:vocab.json",
        "merges": "model:merges.txt",
        "tokenizer_json": "model:tokenizer.json"
      },
      "tensors": {
        "weights": "weights:"
      }
    },
    {
      "format": "safetensors",
      "roots": {
        "model": "."
      },
      "files": {
        "config": "model:config.json",
        "generation_config": "model:generation_config.json",
        "tokenizer_config": "model:tokenizer_config.json"
      },
      "optional_files": {
        "preprocessor_config": "model:preprocessor_config.json",
        "processor_config": "model:processor_config.json",
        "chat_template": "model:chat_template.json",
        "chat_template_jinja": "model:chat_template.jinja",
        "vocab": "model:vocab.json",
        "merges": "model:merges.txt",
        "tokenizer_json": "model:tokenizer.json"
      },
      "tensors": {
        "weights": "model:model.safetensors"
      }
    }
  ],
  "options": {
    "request": [
      {
        "name": "language",
        "type": "string",
        "required": false,
        "default": "Auto",
        "description": "Recognition language: ISO-639 code (zh/en/ja/...) or canonical name (Chinese/English/...), or Auto for detection."
      },
      {
        "name": "max_tokens",
        "type": "int",
        "required": false,
        "description": "Offline decode budget. Must be positive. Offline omission uses checkpoint configuration (512 fallback).",
        "min": 1
      }
    ],
    "session": [
      {
        "name": "chunk_size_ms",
        "type": "float",
        "required": false,
        "default": 320,
        "description": "Streaming chunk duration in milliseconds; must be positive."
      },
      {
        "name": "unfixed_chunk_num",
        "type": "int",
        "required": false,
        "default": 2,
        "description": "Leading chunks decoded without a stable-prefix prompt.",
        "min": 0
      },
      {
        "name": "unfixed_token_num",
        "type": "int",
        "required": false,
        "default": 5,
        "description": "Tokens rolled back from the accumulated text before it becomes the prefix prompt.",
        "min": 0
      },
      {
        "name": "rollback_punctuation",
        "type": "bool",
        "required": false,
        "default": false,
        "description": "When punctuation is detected, set rollback to zero and retain the full decoded prefix."
      },
      {
        "name": "max_tokens",
        "type": "int",
        "required": false,
        "default": 32,
        "description": "Greedy decode budget per streaming chunk and for the final flush (session-scoped; distinct from the offline request max_tokens).",
        "min": 1
      },
      {
        "name": "audio_encoder_weight_type",
        "type": "enum",
        "preset": "weight_type_conv",
        "required": false,
        "default": "native",
        "description": "Audio tower weight storage."
      },
      {
        "name": "thinker_weight_type",
        "type": "enum",
        "values": [
          "native",
          "f32",
          "f16",
          "bf16",
          "q8_0"
        ],
        "required": false,
        "default": "native",
        "description": "Thinker weight storage."
      },
      {
        "name": "weight_type",
        "type": "enum",
        "preset": "weight_type_full",
        "required": false,
        "default": "native",
        "description": "Alias for thinker_weight_type."
      },
      {
        "name": "audio_encoder_graph_arena_mb",
        "type": "int",
        "required": false,
        "default": 128,
        "description": "Audio tower graph arena in MB.",
        "min": 1
      },
      {
        "name": "thinker_prefill_graph_arena_mb",
        "type": "int",
        "required": false,
        "default": 256,
        "description": "Thinker prefill graph arena in MB.",
        "min": 1
      },
      {
        "name": "thinker_decode_graph_arena_mb",
        "type": "int",
        "required": false,
        "default": 256,
        "description": "Thinker decode graph arena in MB.",
        "min": 1
      },
      {
        "name": "thinker_weight_context_mb",
        "type": "int",
        "required": false,
        "default": 64,
        "description": "Thinker weight context in MB.",
        "min": 1
      }
    ],
    "load": []
  },
  "dependencies": []
}