Belle-whisper-large-v3-zh β€” MLX bfloat16

This is an MLX-format conversion of BELLE-2/Belle-whisper-large-v3-zh in bfloat16, suitable for Apple Silicon inference via mlx-audio and downstream tools like oMLX.

Conversion

Produced from the upstream PyTorch checkpoint with mlx-audio 0.4.3:

python -m mlx_audio.convert \
  --hf-path BELLE-2/Belle-whisper-large-v3-zh \
  --mlx-path ./Belle-whisper-large-v3-zh-mlx-bf16 \
  --dtype bfloat16 \
  --model-domain stt

The output directory contains the full HF processor surface (tokenizer.json, preprocessor_config.json, generation_config.json, etc.), unlike older legacy MLX Whisper conversions that only shipped config.json + weights.safetensors and could not be loaded by modern WhisperProcessor-based stacks. If you have a legacy directory to upgrade in place, see mlx-whisper-legacy-fixup.

Usage

Via mlx-audio

from mlx_audio.stt.generate import generate_transcription

result = generate_transcription(
    model="BRlin/Belle-whisper-large-v3-zh-mlx-bf16",
    audio="path/to/audio.wav",
    language="zh",
)
print(result.text)

Via oMLX HTTP endpoint

curl -X POST http://127.0.0.1:8868/v1/audio/transcriptions \
  -H "Authorization: Bearer $OMLX_KEY" \
  -F "file=@path/to/audio.m4a" \
  -F "model=Belle-whisper-large-v3-zh-mlx-bf16" \
  -F "language=zh"

Specs

Field Value
Architecture Whisper large-v3 (fine-tuned)
Parameters ~1.55 B
Precision bfloat16
n_mels 128
Vocab size 51866
File size ~2.9 GB
Sample rate 16 kHz mono
Best for Mandarin Chinese ASR (Simplified Chinese output)

License & Attribution

Inherits Apache 2.0 from the upstream BELLE-2 model. All credit for the underlying weights belongs to the BELLE-2 / LianjiaTech team and the Whisper-Finetune project β€” this repository only provides the MLX format conversion for convenience on Apple Silicon.

If you use this model, please cite the upstream:

Maintenance

This is a one-shot conversion at 2026-04 and is not actively maintained beyond bug fixes. If the upstream BELLE-2 model is updated, please re-run the conversion command above, or open an issue.

Downloads last month
13
Safetensors
Model size
2B params
Tensor type
F16
Β·
MLX
Hardware compatibility
Log In to add your hardware

Quantized

Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support

Model tree for BRlin/Belle-whisper-large-v3-zh-mlx-bf16

Finetuned
(8)
this model