OmniVoice MLX q4-g64: Qwen3 backbone quantized via mlx-lm; Higgs tokenizer bf16

c2c7cf4

verified

lightsofapollo commited on 7 days ago