Post-Cutoff

Multimodal Art Projection (M-A-P)YuE familyCurrentMusic generationReleased 9 Sep 2026

YuE2-3B

Open-weights full-song generator from M-A-P (HKUST-led). Runs locally on one consumer GPU; weights download automatically on first use via the GitHub package.

PricePrice not published

Checked29 September 2026

Where it stands

1newer music generation model since its release

All music generation models

Weights
Weights are public. Licence: cc-by-nc-4.0, apache-2.0. See the licence for commercial terms.

How to call it

3 ways to use YuE2-3B.

Providers, model ids, endpoints and docs for YuE2-3B
ProviderModel idEndpointDocs
Hugging Face huggingface.co
GitHub (inference code, agent skill) github.com
Hugging Face (community GGUF) huggingface.co

Pricing

Price not published.

No price source linked. Checked 29 September 2026.

Specs

Input
Text and audio
Output
Audio
Context window
Not published
Max output
Not published
Open weights
Yes
Licence
cc-by-nc-4.0 (weights; commercial license available), apache-2.0 (code)
Released
9 September 2026
Status
Current
Type
Music generation

What stands out

What its maker marketed at launch, and what people found later, each with its source.

  • Score-first song generation

    Writes an editable melody-and-chord plan in ABC notation, then renders a full song with vocals and accompaniment (48 kHz stereo).

    Source: github.com

  • Zero-shot covers and agentic editing

    Covers from reference recordings (0.647 CLEWS mAP, self-reported) and conversational editing that turns musical feedback into score revisions.

    Source: huggingface.co

Notes

Self-reported WildSongBench best-of-8 6.9632 vs Suno v5 6.8721. English + Mandarin lyrics. Model card states ~4B parameters. Predecessor YuE (2025-01-28, arXiv 2503.08638).

Sources: https://github.com/multimodal-art-projection/YuE , https://huggingface.co/m-a-p/YuE2-3B

Not yet checked against an official source: HF repos created 2026-09-09; exact public announcement day not verified.