For most of 2026, Suno has been the name to beat in AI song generation. This week a smaller rival edged past it on the one independent listening test the category has. ACE Music v1, the first full-song model from ACE Studio, now ranks first on Artificial Analysis’s AA-Music-Vocal v1.1 leaderboard, with an Elo score of 1,160 against Suno v6’s 1,144.
It is a narrow win, and the details matter more than the headline. Here is what the leaderboard actually measures, how close the race really is, what ACE Music v1 can do, and what it means if you make or listen to AI-assisted music.
What the leaderboard shows
Artificial Analysis is an independent company that benchmarks AI models. For music, it runs two boards: one for songs with vocals and one for instrumentals. On the vocal board, as of 12 November 2026, the top five are ACE Music v1 (1,160), Suno v6 (1,144), Suno v6-mini (1,132), Mureka V9.5 (1,101) and Mureka V9 (1,063). Google’s Lyria 3.5 sits sixth on 1,050, and ElevenLabs’ Eleven Music v2.5 is eighth on 1,016.
The method is simple. Each model generates a track from the same prompt. Validated private evaluators then listen to pairs of tracks without knowing which model made which, and pick the one they prefer. Those votes become Elo ratings, the same kind of system used to rank chess players. MiniMax Music 3.0 is fixed at 1,000 as the reference point, so every other score is relative to it.
How close is the race?
Very close. Artificial Analysis publishes a 95% confidence interval for each model. ACE Music v1’s runs from 1,139 to 1,181, and Suno v6’s runs from 1,126 to 1,162. Because those ranges overlap, the leaderboard itself lists ACE’s possible rank as first or second, and Suno v6’s as anywhere from first to third.
A 16-point Elo gap also translates into a modest edge in practice. Under the standard Elo formula, it works out to the higher-rated model being preferred roughly 52% of the time in a head-to-head. ACE Music v1 has also been judged in fewer comparisons so far, 2,119 against Suno v6’s 3,063, so its score could still move as more votes come in.
Suno keeps a clear lead elsewhere. On Artificial Analysis’s separate AA-Music-Instrumental v1.1 board, Suno v6 is first with 1,148 and Suno v6-mini second with 1,114. ACE Music v1 does not appear on that board at all, which fits its roots in vocals.
Who is ACE Studio?
ACE Studio was founded in 2018 and started out synthesising singing voices from MIDI notes and lyrics, Decrypt reports. Its new ACE Music v1 model goes further, generating complete songs from a written description or a user’s own lyrics. ACE Studio has also co-developed ACE-Step, a separate family of open-source music models, with the AI company StepFun.
According to ACE Music’s own product page, you can add performance directions in brackets inside your lyrics, such as “hum softly” or “sing with a raspy voice”, to shape how a line is delivered. After generating a song, you can convert it to MIDI to keep arranging in a DAW and split it into stems for mixing. A second model, ACE Music v1.1 Preview, plans the music as a score in ABC notation before generating audio, so you can edit the melody and regenerate.
“We plan to open-source ACE Music v1 Flash in late 2026.” — ACE Music
ACE says the Flash version will trade some musical quality for speed and lower hardware requirements, making it practical to run locally. The flagship ACE Music v1 has no public API, so for now you can only use it through ACE Music’s website.
What a hands-on test found
Decrypt put the two models side by side. Its reviewer found ACE Music consistently produced more melodic and coherent songs in mainstream styles, with more natural-sounding vocals and cleaner endings. Suno, it reported, sometimes adds odd sounds, repetitions or abrupt changes near the end of a track.
Suno still won some rounds. In Decrypt’s melodic death metal test, ACE kept producing a clean, pleasant female vocal instead of the aggressive delivery the genre calls for, and Decrypt said ACE “confuses metal genres all over”. Suno handled heavier and less conventional styles with more creative and consistent arrangements. Decrypt was clear that its tests were subjective, not a controlled benchmark.
Rights and commercial use
If you plan to release what you make, the plan you are on matters. ACE Music says songs generated with ACE Music v1 on a paid plan come with commercial use rights, and those rights stay if you later cancel. Songs made on the free plan do not, and upgrading later does not change that. Always read a tool’s terms before you distribute a track.
The legal backdrop is still unsettled for the whole category. Suno faces lawsuits from Universal Music Group and Sony Music, and new suits this month from Elvis Presley rights holders and Eminem’s publisher, which we covered in Elvis rights holders and Eminem’s publisher file new lawsuits against Suno. Those claims are allegations that have not been decided in court.
Why it matters
For creators
Competition is good news if you make music with AI. A credible challenger pushes every tool to improve vocals, structure and editing control, and features such as MIDI export, stems and editable scores give you more of a say in the finished song. The practical lesson from this leaderboard is that no single model wins at everything. Try the same idea in more than one tool, and pick the one that suits your genre.
For listeners
Blind preference tests are a useful reminder that listeners judge the song, not the logo. As vocal quality improves across tools, what makes a track stand out is less about which model made it and more about the ideas, lyrics and choices the creator brings. That is exactly what charts like Dizy’s are designed to surface.
What happens next
Watch whether ACE Music v1 holds first place as its sample size grows, whether Suno answers with a new release, and whether ACE ships the open-source Flash model before the end of 2026 as promised.
Hear the best AI-assisted songs on Dizy
Whether you make music with ACE, Suno or another tool, Dizy charts original AI-assisted songs from independent creators. Explore the Top 40, the Top 100 and New releases, discover AI artists, or submit your track to see where it lands.
Sources
Artificial Analysis: AA-Music-Vocal v1.1 leaderboard (accessed 12 November 2026)
Artificial Analysis: AA-Music-Instrumental v1.1 leaderboard (accessed 12 November 2026)
