{
  "model_type": "minimax_music3",
  "quant": "8bit",
  "group_size": 64,
  "frame_rate": 25.0,
  "latent_rate": 86.1328125,
  "sample_rate": 44100,
  "max_frames": 9000,
  "max_text_tokens": 5000,
  "num_codebooks": 8,
  "semantic_vocab_size": 16384,
  "audio_vocab_size": 1024,
  "language_model": {
    "hidden_size": 4096,
    "num_hidden_layers": 36,
    "num_attention_heads": 32,
    "num_key_value_heads": 8,
    "head_dim": 128,
    "intermediate_size": 12288,
    "vocab_size": 200000,
    "rms_norm_eps": 1e-06,
    "rope_theta": 1000000,
    "max_position_embeddings": 10240
  },
  "rvq_depth_decoder": {
    "hidden_size": 4096,
    "intermediate_size": 6144,
    "num_attention_heads": 16,
    "num_layers": 4,
    "max_position_embeddings": 16
  },
  "transformer": {
    "num_layers": 36,
    "num_attention_heads": 32,
    "attention_head_dim": 64,
    "hidden_size": 2048,
    "ff_inner_dim": 8192,
    "in_channels": 128,
    "condition_dim": 2048,
    "fourier_embedding_dim": 256,
    "rotary_dim": 32
  },
  "condition_encoder": {
    "condition_hidden_dim": 4096,
    "out_dim": 2048,
    "num_condition_layers": 8,
    "input_hop_length": 960,
    "input_sampling_rate": 24000,
    "output_hop_length": 512,
    "output_sampling_rate": 44100
  },
  "vocoder": {
    "latent_channels": 128,
    "decoder_input_dim": 1024,
    "decoder_hidden_dim": 1536,
    "upsampling_ratios": [
      8,
      8,
      4,
      2
    ]
  },
  "scheduler": {
    "type": "flow_match_euler_discrete",
    "num_train_timesteps": 1,
    "shift": 1.0,
    "invert_sigmas": true
  }
}