ACE Music v1 takes first place on Artificial Analysis's vocal music board, inside the margin over Suno v6

ACE Studio’s new song model, ACE Music v1, is first on Artificial Analysis’s AA-Music-Vocal v1.1 leaderboard, the board for generated music with singing. It turns a text description or your own lyrics into a full song in the browser.

What the leaderboard says

Artificial Analysis ranks music models with Elo scores from blind pairwise votes: validated evaluators hear two tracks made from the same prompt and pick one without knowing the models. On 12 November the top of the vocal board read:

Rank range	Model	Elo	95% interval	Samples
1-2	ACE Music v1	1160	1139 to 1181	2,119
1-3	Suno v6	1144	1126 to 1162	3,063
2-3	Suno v6-mini	1132	1114 to 1150	3,175
4	Mureka V9.5	1101	1083 to 1119	3,141

The intervals for ACE Music v1 and Suno v6 overlap, and Artificial Analysis itself gives ACE a rank range of 1-2 and Suno v6 a range of 1-3. So “first” is accurate, “clearly better than Suno” is not. ACE’s own FAQ is careful too: the ranking reflects blind preferences for vocal music, “not separate rankings for every genre or feature.” Artificial Analysis keeps a separate instrumental board, where Decrypt reports Suno still leads.

What you can do with it

• Text or lyrics in, song out. You can add performance cues in parentheses inside the lyrics, such as “hum softly” or “add guitar here”.
• After generation, ACE Music exports MIDI for a DAW and splits stems for vocals and instruments.
• ACE Music v1.1 Preview is a separate model that generates from melodies and scores, planning the music in ABC notation first.
• Rights. ACE’s page says songs made on a paid plan have commercial use rights; songs made on the free plan do not, and upgrading later does not grant them.
• No API or open weights today. It runs in ACE’s browser app. ACE says it plans to open-source a lighter ACE Music v1 Flash in late 2026 for local generation.

Decrypt’s hands-on review found stronger melodies, more natural voices and cleaner song endings than Suno, with Suno better on some niche genres. That is one outlet’s listening test, not a measurement.

Why this matters

If you choose a music model from a leaderboard, the top three on this board sit inside each other’s error bars, so test with your own prompts and genres before switching. ACE Music v1 is not something you can call from code today; for builders the item to watch is the promised open Flash model. Check the commercial-rights terms of any music generator before you publish, especially as platforms such as YouTube Music begin labelling primarily AI-made tracks. Artificial Analysis has also moved its text model index toward private test sets; its music boards rest on blind human votes instead, which is why the intervals matter.