---
license: other
license_name: minimax-music3-community-license
license_link: LICENSE
base_model: MiniMaxAI/MiniMax-Music3
base_model_relation: quantized
library_name: mlx-serve
tags:
  - mlx
  - mlx-serve
  - music
  - text-to-music
  - minimax
pipeline_tag: text-to-audio
---

# MiniMax Music 3 for mlx-serve (8-bit)

[MiniMax-Music3](https://huggingface.co/MiniMaxAI/MiniMax-Music3) converted for
[mlx-serve](https://github.com/ddalcu/mlx-serve)'s native Zig + MLX engine.
Full songs with sung lyrics at 44.1 kHz stereo, generated locally on Apple
Silicon.

Quantization: affine 8-bit, group 64, on every real matmul (LLM, depth
decoder, DiT, lm_head). Kept dense on purpose: the embedding tables (gather
reads), the condition encoder and the whole vocoder (VAE-class precision).
Worst per-tensor reconstruction error 1.48% RMS. 13 GB on disk instead of the
57 GB upstream repo (which ships its weights twice).

| File | Contents |
|---|---|
| `language_model.safetensors` | Qwen3 8B global LLM, 36L, vocab 200k |
| `rvq_depth_decoder.safetensors` | 0.6B local LLM, 7 residual codebooks |
| `transformer.safetensors` | 2.4B flow-matching DiT, 36 blocks |
| `condition_encoder.safetensors` | hidden-state mix + resampler, f32 |
| `vocoder.safetensors` | Flow-VAE / DAC decoder, f32 |

Engine parity against the fp32 reference on these exact weights: prefill
cosine 0.9999, condition encoder 0.999999, DiT velocity 0.999, vocoder
1.000000. The autoregressive stage runs about 44 ms per frame on an M-series
Mac, so a one-minute song takes roughly a minute of LLM time plus the
diffusion pass.

## Run it

Download **[MLX Core.app](https://github.com/ddalcu/mlx-serve/releases/latest)**,
open the Music tab and pick **MiniMax Music 3**. Style prompt + lyrics in,
WAV out.

Over HTTP:

```bash
mlx-serve --serve
curl http://127.0.0.1:11234/v1/audio/music-generations \
  -H 'Content-Type: application/json' -o song.wav -d '{
    "model": "MiniMax-Music3-MLX-Serve-8bit",
    "prompt": "upbeat synthwave with driving bass and dreamy pads",
    "lyrics": "[verse]\nneon lights across the bay\n[chorus]\nwe run all night",
    "duration_seconds": 60
  }'
```

Lyrics are required (the model is lyric-conditioned) and structure tags like
`[verse]` or `[chorus]` go on their own lines. `duration_seconds` (1-360) is
an upper bound, the model may end the song earlier. ACE-Step style fields
(bpm, keyscale, timesignature, vocal_language) do not exist on this model.

Rebuild from the upstream repo with
[`scripts/convert_music3_weights.py`](https://github.com/ddalcu/mlx-serve/blob/main/scripts/convert_music3_weights.py).

Weights are covered by the MiniMax-Music3 Community License (see `LICENSE`,
acceptable-use policy included as Exhibit A).