HKUST-led team open-sources YuE2 music model that beats Suno v6 on WildSongBench
量子位 · wechat · 2026-09-20
HKUST with M-A-P, NOIZAI and Stanford released YuE2, an open-source 3.6B-parameter music generation model (Mixture-of-Transformers) that scores 6.7316 on WildSongBench across 192 prompts—closing in on Suno v5 and Mureka 9 while beating Suno v6/v6Wild, and topping 17 systems at 6.9632 in pick-one-of-eight settings.
Its key innovation is symbolic planning: the model first writes an editable ABC-notation lead sheet (vocal/instrumental melodies, chords, structure) before rendering audio. This enables Agentic Music Editing—in targeted edit tests, target pitch match rose from 0.83% to 93.75% and chord consistency from 0 to 83.13% while untouched parts stayed above 0.96/0.93 similarity. CoverSong regeneration via SheetSage2 hit 71.3% top-1 retrieval with full sheets. The release also includes the MERT2 music-understanding model (leading on 14/15 MARBLE metrics). Runs on a single RTX 4090, generating a song in a minute. Weights on GitHub and Hugging Face.
More from Multimodal
- Izwi: open-source offline voice cloning in Rust, runs on 16GB M1 Pro — zinyando · 2026-11-12
- MiniMax H3 Image generates 3D game scenes in 10s, over 3x faster than GPT Image — HeyToha · 2026-11-12
- Dev releases local world model turning any image into interactive video, runs on RTX 3060 — lucidml_lover · 2026-11-12
- Ex-Meta CV researcher builds AdaL v2, a chat-driven video agent harness for GTM — Scobleizer · 2026-11-12
- MiniMax H3 image gen runs locally on AMD RX 9070, 25-30s per 4MP image — omglionheaded · 2026-11-12
- One reference, one prompt, three image models: side-by-side Nano Banana Pro vs GPT Image 2.6 test — gizakdag · 2026-11-12