Skip to content
localmodel.run

Audio model · musicgen

MusicGen medium requirements

Music generation · 1.5B params · fp32 · released Jun 2023. Realistic minimum to run: Nvidia GeForce RTX 4060 Ti (16GB).

CC-BY-NC-4.0 Non-commercial

Model weights are CC-BY-NC 4.0 (non-commercial); the AudioCraft code is MIT.

Peak memory (fp32)
~14 GB
Runs on CPU
No
Parameters
1.5B
Type
Music generation

Run it

Runtime tools fp32

MusicGen medium runs in AudioCraft or HF Transformers at fp32. It needs a GPU (or Apple Silicon with enough unified memory).

AudioCraftHF Transformers

Which devices can run MusicGen medium?

Apple Silicon Macs

No mainstream local runtime for MusicGen medium on Apple Silicon Macs yet.

RAM-only laptops

No mainstream local runtime for MusicGen medium on RAM-only laptops yet.

iPhone & iPad

No mainstream local runtime for MusicGen medium on iPhone & iPad yet.

Android

No mainstream local runtime for MusicGen medium on Android yet.

NVIDIA GPUs

AMD GPUs

FAQ

How much memory does MusicGen medium need?

At fp32 it consumes ~14 GB. It needs a GPU or Apple Silicon.

Can MusicGen medium run on a phone or CPU?

Not practically. MusicGen medium expects a GPU; CPU inference is too slow for real use.

Can I use MusicGen medium commercially?

No. Model weights are CC-BY-NC 4.0 (non-commercial); the AudioCraft code is MIT.

Notes

MusicGen medium (1.5B). AudioCraft's docs state it requires a GPU with at least 16GB of memory; peak consumed is ~14GB fp32 (synthesis from the official 16GB-GPU guidance and the 1.5B weight size). Weights are CC-BY-NC (non-commercial). Sources: HF card, AudioCraft MusicGen docs.

Sources

Memory is a sourced peak-usage anchor at fp32 (composed from reported sizes, not a single measurement), validated 2026-06-15. See methodology.