YuE2-3B Music Generation Model for Lyrics and Style: 12GB+ VRAM
Sample outputs are available on the model card.
At a Glance
| Item | Value |
|---|---|
| Repository | m-a-p/YuE2-3B |
| Family guide | YuE2 guide (3 articles) |
| Publisher guide | M-A-P: models and licenses |
| Published | 2026-09-10 |
| License | cc-by-nc-4.0 |
| Formats | safetensors |
| Paper | arXiv:2503.08638 |
| Source type | Primary source (the publisher itself) |
Values determined by this site’s code when the information was collected. Dates are JST.
Overview
m-a-p has released “YuE2-3B", an open music generation model that generates complete songs with vocals and accompaniment from lyrics and style prompts. It adopts a mechanism where an AR-NAR Mixture-of-Transformers backbone describes scores and semantic tokens, acoustic latent representations are generated via flow matching, and a VAE converts them into stereo audio.
Specifications
- Architecture: AR-NAR Mixture-of-Transformers (YuE2ForCausalLM)
- Output Specification: 48 kHz stereo audio
- Supported Languages: Chinese (zh), English (en)
- License: cc-by-nc-4.0 (non-commercial use only)
- Number of Parameters: 3.6B
Performance and Quality
Evaluation results on WildSongBench and SHS100K are published in the model card.
→ Scroll horizontally to see all columns
| Model | Musicality ↑ | SongBench Avg ↑ | MuLan ↑ | AllMusicCaps ↑ | Q3O ↑ | PER ↓ |
|---|---|---|---|---|---|---|
| YuE 1 | 4.0847 | 4.9165 | 0.2623 | 0.2882 | 3.7301 | 36.38% |
| SongBloom | 3.4493 | 4.2350 | 0.2697 | 0.1926 | 3.0287 | 19.19% |
| LeVo 2 | 5.4590 | 6.3247 | 0.3542 | 0.2680 | 3.9458 | 26.12% |
| ACE-Step 1.5 | 5.1588 | 6.0118 | 0.4372 | 0.3869 | 4.5809 | 7.46% |
| HeartMuLa | 5.4963 | 6.2483 | 0.3823 | 0.2786 | 3.4907 | 10.71% |
| DiffRhythm 2 | 4.4775 | 5.2428 | 0.3782 | 0.3255 | 4.0870 | 18.41% |
| Muse | 5.1692 | 6.0349 | 0.3937 | 0.3466 | 4.4038 | 33.42% |
| MiniMax Music 3 | 5.3482 | 6.2830 | 0.3928 | 0.3609 | 4.4362 | 6.27% |
| YuE2 | 5.9075 | 6.7316 | 0.5068 | 0.4054 | 4.6819 | 8.44% |
| YuE2 (best-of-8) | 6.2666 | 6.9632 | 0.5051 | 0.3980 | 4.7009 | 9.79% |
→ Scroll horizontally to see all columns
| Model | Musicality ↑ | SongBench Avg ↑ | MuLan ↑ | AllMusicCaps ↑ | Q3O ↑ | PER ↓ |
|---|---|---|---|---|---|---|
| Suno v5 | 5.9918 | 6.8721 | 0.5428 | 0.4353 | 4.5907 | 8.10% |
| Suno v4.5 | 5.8317 | 6.6995 | 0.5022 | 0.3873 | 4.4149 | 5.80% |
| Suno v5.5 | 5.8087 | 6.7150 | 0.5089 | 0.3917 | 4.5914 | 5.96% |
| MiniMax Music 2.6 | 5.4437 | 6.3222 | 0.4251 | 0.3670 | 4.5688 | 24.55% |
| Mureka 9 | 6.0488 | 6.9377 | 0.4394 | 0.4102 | 4.6368 | 11.69% |
| YuE2 | 5.9075 | 6.7316 | 0.5068 | 0.4054 | 4.6819 | 8.44% |
| YuE2 (best-of-8) | 6.2666 | 6.9632 | 0.5051 | 0.3980 | 4.7009 | 9.79% |
In WildSongBench evaluations, it is reported that YuE2 (best-of-8) achieved a SongBench average of 6.9632, the highest value among the compared open and proprietary models. On the other hand, it shows lower scores than some models in metrics such as PER (lower is better).
Strengths and Use Cases
- Text-to-song generation (complete music production via lyrics and style prompt input)
- Cover production of existing songs (reconstructing existing tracks into new styles)
- Editable scores (composition and editing by bringing melody and chords, melody only, or ABC scores)
- Agent-leveraged editing (reflecting musical feedback into revisions of scores, styles, and lyrics to render the next version)
Hardware Requirements
Estimated requirements (calculated by Local Model Watch) — 3.6B parameters
| Your VRAM | Quantization | File size | Est. memory needed |
|---|---|---|---|
| 12GB (RTX 4070 / 3060 12GB, etc.) | BF16 | 6.8GB | 8.1GB |
Inference engine support (architecture name matched against each project’s own model registry in its source code, checked 2026-11-03): llama.cpp: not registered, vLLM: not registered, MLX (mlx-lm): not registered. “Not registered" means the name is absent from that registry today, not that the model cannot run.
Memory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference. What the quantization names mean: glossary.
Can You Run It Locally?
The publisher distributes this model as safetensors.
License — cc-by-nc-4.0 (Commercial use prohibited):
Commercial use is prohibited (NC = NonCommercial). Internal business use can also count as commercial, so avoid this license for work use.
Compiled by this site’s code from the published formats and the license field. License summaries are not legal advice — check the publisher’s original terms before relying on them.
Our Own Measurements
Measurements We Did Not Take
This model’s license (cc-by-nc-4.0) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).
Distributed Files
Weight files published in m-a-p/YuE2-3B, listed by this site from the Hugging Face API. Sizes are the actual file sizes.
| File | Size |
|---|---|
model.safetensors |
7.26GB |
How to Get It
- Distribution Format: safetensors
- Source and Related Repository:
m-a-p/YuE2-3Bon Hugging Face
Related Articles
What to Read Next
- Find models by VRAM (This model runs from the 12GB tier) → Other models that run on a 12GB GPU
- Explore the same model family → YuE2 family overview (3 articles)
- Formats this model is available in → Safetensors format guide and models
- Learn about the publisher → M-A-P: models, licenses and articles
- Other models for the same task → Other audio models
Sources
Update History
- 2026-09-24: Rebuilt the article: Added a “can you run it locally?" verdict from the distribution formats, converted builds, engine registries, license terms and measured compression
- 2026-09-25: Changed the title to show what the article covers (VRAM requirements, file list, etc.).