ACEMusic v1 tops Artificial Analysis vocal chart, dethroning Suno v6 for the first time

机器之心 · wechat · 2026-11-20

ACE's ACEMusic v1 scored 1160 Elo on the Artificial Analysis vocal leaderboard, beating Suno v6 (1144) — the first time a Suno flagship has been dethroned since the chart launched. ACE took only 18 months from starting R&D to the top spot.

AA uses blind human preference voting. Hands-on tests show ACEMusic handles diverse styles (funk, K-pop, orchestral) with stable arrangement quality, plus image-to-song, hum-to-song, and one-click cover features.

Its core technique is Symbolic CoT: an LLM first generates an explicit musical score, then a music model renders it — enabling score-level editing, direct use of LLM music knowledge, and agentic workflows (e.g., Claude autonomously arranging a 5-minute, 88-bar Mandarin pop medley). ACE also runs ACE Studio for pros with 3M+ creative trajectories feeding a data flywheel, and raised $40M in September 2026.

Related event: ACE Music v1 tops vocal music leaderboard, beating Suno(3 posts)→

Original post →

More from Multimodal

Multimodal channel →