---
title: MiniMax-Music3-bf16
canonical_url: "https://www.modelscope.cn/models/mlx-community/MiniMax-Music3-bf16"
md_url: "https://www.modelscope.cn/models/mlx-community/MiniMax-Music3-bf16.md"
repository: mlx-community/MiniMax-Music3-bf16
last_updated: 2026-08-17
license: other
model_type:
  - minimax_music3
architectures:
  - MiniMaxMusic3ForConditionalGeneration
base_model:
  - MiniMaxAI/MiniMax-Music3
base_model_relation: finetune
parameters: 11.7B
tensor_type:
  - BF16
  - F32
library_name:
  - mlx
  - safetensors
  - pytorch
frameworks:
  - pytorch
downloads: 57
stars: 0
tags:
  - mlx
  - mlx-audio
  - apple-silicon
  - music-generation
  - text-to-music
  - text-to-audio
  - bf16
---

# MiniMax-Music3-bf16

> MiniMax-Music3-bf16 - mlx-community 在 ModelScope 开源的模型。MiniMax Music 3 · MLX BF16

mlx-community/MiniMax-Music3-bf16 是 ModelScope 魔搭社区上的 11.7B 参数机器学习模型，采用 other 许可，基于 MiniMaxAI/MiniMax-Music3 构建。

- **Repository**: mlx-community/MiniMax-Music3-bf16
- **License**: other
- **Parameters**: 11.7B
- **Base model**: MiniMaxAI/MiniMax-Music3
- **Tags**: mlx, mlx-audio, apple-silicon, music-generation, text-to-music, text-to-audio, bf16
- **Downloads**: 57
- **Stars**: 0
- **Last updated**: 2026-08-17

Source: https://www.modelscope.cn/models/mlx-community/MiniMax-Music3-bf16

---

# MiniMax Music 3 · MLX BF16

Native MLX BF16 weights for
[`MiniMaxAI/MiniMax-Music3`](https://huggingface.co/MiniMaxAI/MiniMax-Music3),
converted for lyric-conditioned song generation on Apple Silicon with
[`mlx-audio`](https://github.com/Blaizzy/mlx-audio).

> Community conversion, not an official MiniMax release. All model credit goes
> to MiniMax. Review the original model card and license before use.

Other MLX variants:
[`8-bit`](https://huggingface.co/mlx-community/MiniMax-Music3-8bit)
· [`6-bit`](https://huggingface.co/mlx-community/MiniMax-Music3-6bit)
· [`4-bit`](https://huggingface.co/mlx-community/MiniMax-Music3-4bit)
· [`MXFP8 (recommended quantized)`](https://huggingface.co/mlx-community/MiniMax-Music3-mxfp8)
· [`MXFP4 (experimental)`](https://huggingface.co/mlx-community/MiniMax-Music3-mxfp4)
· [`NVFP4 (experimental)`](https://huggingface.co/mlx-community/MiniMax-Music3-nvfp4)

## Install

MiniMax Music 3 support was merged upstream in
[`Blaizzy/mlx-audio#888`](https://github.com/Blaizzy/mlx-audio/pull/888).
Until a PyPI release includes it, install the upstream merge commit directly:

```bash
python -m pip install "mlx-audio @ git+https://github.com/Blaizzy/mlx-audio.git@784b29e2691a93ca7483147d86f61859dfaa6296"
```

## Generate

```bash
python -m mlx_audio.music.generate \
  --model mlx-community/MiniMax-Music3-bf16 \
  --caption "Warm acoustic pop, 96 BPM, intimate female vocal" \
  --lyrics $'[verse]\nMorning light across the room\n[chorus]\nSing with me' \
  --duration 30 \
  --steps 30 \
  --seed 7 \
  --output song.wav
```

```python
from mlx_audio.music import load

model = load("mlx-community/MiniMax-Music3-bf16")
result = next(
    model.generate(
        text="Warm acoustic pop, 96 BPM, intimate female vocal",
        lyrics="[verse]\nMorning light across the room\n[chorus]\nSing with me",
        duration=30,
        steps=30,
        seed=7,
    )
)
print(result.audio.shape, result.sample_rate)  # stereo, 44100 Hz
```

Lyrics are required by the checkpoint contract. Use `[instrumental]` explicitly
for instrumental generation. Duration is a requested upper bound: the
autoregressive stage may emit its end token early. Style, tempo, instrument, and
vocal controls are probabilistic rather than strict.

## Conversion and verification

- Dense BF16; no weight quantization.
- Complete native pipeline: Qwen3 global autoregressive model, RVQ depth
  decoder, condition encoder, flow-matching DiT/Euler stage, and stereo vocoder.
- The actual component configs produce 44.1 kHz stereo output.
- Official-checkpoint conversion maps and strict-loads all 982 expected tensors.
- Full-checkpoint float32 maximum absolute differences against the official
  PyTorch implementations: Qwen logits `1.45e-4`, RVQ depth `2.50e-5`, condition
  encoder `1.91e-6`, flow transformer `7.63e-6`, vocoder `8.57e-7`.
- The mlx-audio regression suite passed 1,742 tests with 34 expected skips; the
  focused music, converter, and registry suite passed 43 tests and 3 subtests.
- Real generation produced finite 44.1 kHz stereo audio. A 210-second request
  exercised 38 denoising windows and ended at 152.8 seconds when the model
  emitted EOS.

Converted with `mlx-audio` 0.4.8 development commit `c2fa486` and MLX 0.31.2.

## License

The weights remain subject to the
[`MiniMax-Music3 Community License`](https://huggingface.co/MiniMaxAI/MiniMax-Music3/blob/main/LICENSE),
including its acceptable-use and commercial terms. The full license text is
included in this repository.