Performance
On a Mac
One full song on a Mac mini (M4, 16 GB), macOS 26.5, MLX 0.32.2. The model ended the song on its own after 61.9 seconds of audio.
| Stage | Seconds |
|---|---|
| Load (tokenizer, AR, NAR) | 1.7 |
| Score (ABC), 691 tokens at 80 tokens/s | 9.6 |
| Codec tokens, 1,548 at 85 tokens/s | 19.0 |
| NAR (codec → latents), 32 steps | 116.1 |
| VAE decode | 5.2 |
| Total | 152 |
Peak memory: 3.5 GB, during VAE decode.
On iPhone
A 30-second clip on an iPhone 17 Pro Max, iOS 26.6.
| Stage | Seconds |
|---|---|
| Load (tokenizer, AR) | 0.6 |
| Score (ABC), 26 tokens/s | 23.2 |
| Codec tokens (750), 41 tokens/s | 18.8 |
| NAR (codec → latents) | 34.3 |
| VAE decode | 4.3 |
| Total | 81.6 |
Peak memory: 1.9 GB while generating, 3.6 GB during VAE decode. The phone stayed at a nominal thermal state the whole run.