ACE Music v1 tops the AI music leaderboard ahead of Suno v6: my test

👁 111 reads

On the morning of 11 November 2026 Artificial Analysis put a new name at the top of its AI music leaderboard. ACE Music v1 from ACE Studio debuted at number one for songs with vocals, with 1160 Elo points, pushing Suno v6 to second place. The same morning I made 28 tracks with it on a paid account. In my lyric-accuracy test it sang 104 words out of 104 without a single error, the best result I have measured in any song generator so far.

The short version: it is fast (two tracks in 35-50 seconds), cheap in credits and very clear with lyrics. The weak spots are rap, tongue twisters and the lack of an API. Everything about the model, the plans and the free tier is in my ACE Music guide.

The leaderboard: ACE Music v1 ahead of Suno v6

Artificial Analysis plays pairs of tracks generated from the same prompt to a closed panel of listeners, without model names, and turns thousands of choices into a chess-style rating. MiniMax Music 3.0 is the anchor at 1000.

AA-Music-Vocal v1.1, read on 11 November 2026.

Two things the headlines skip. The lead over Suno v6 is 16 points and the confidence intervals overlap (ACE 1139-1181, Suno 1126-1162), so Artificial Analysis gives ACE a rank range of 1-2. ACE also has the fewest ratings on the board (2,119 against more than 3,000 for Suno). And on the instrumental leaderboard ACE Music v1 does not appear at all: Suno v6 still leads there with 1148.

According to Artificial Analysis the model is strongest in reggae and dancehall and in classical and orchestral music, and weaker in dense modern pop and deliberately odd genre mixes.

Who is ACE Studio

The company was founded in 2018 and started with AI singing synthesis. Its ACE Studio app is a workstation for musicians: you enter MIDI notes and lyrics and it sings them. Together with StepFun, ACE Studio also builds ACE-Step, a family of open-weight music models. ACE Music v1 is different: a closed model that runs only in the acemusic.ai app.

My test: 28 tracks on launch day

I tested it on 11 November on the Artist plan. I wrote all lyrics myself and deliberately packed them with hard consonant clusters (I use Polish lyrics for this, the hardest case I know). Every vocal was separated from the backing track (Demucs) and transcribed by Whisper large-v3, and I counted the words sung without error.

• Reference ballad test: ACE Music v1 sang 104 of 104 words in two takes, the new v1.1 Preview 103 of 104. In the same test in October the best scores were 103 (Eleven Music v2.5 and Lyria 3 Pro).
• Six genres plus the reference test (pop, rap, a dance genre, ballad, rock, reggae): 732 of 762 words, or 96.1%. Ballad, rock and reggae 99%, pop 94%, rap 90%.
• English reggae, one of the genres Artificial Analysis praises: 55 and 56 of 58 words.
• Speed: every generation returns two variants in 35-50 seconds. A track set to 3 minutes came out at 179 seconds in 65 seconds.
• Cost: 4 credits per generation, so 2 credits per track, regardless of length. 28 tracks used 56 credits.

Reference ballad, ACE Music v1, take 1: 52 of 52 words (first 60 seconds). Generated with ACE Music v1 (author’s account).

English roots reggae: 55 of 58 words (first 60 seconds). Generated with ACE Music v1 (author’s account).

Where it slips: tongue twisters and fast rap. With automatic length, 12 lines of rap came out as tracks of 31 and 42 seconds, and the model dropped single sounds in the rush. Set the length by hand (Advanced, Duration) for rap.

Where it runs and what it costs

ACE Music v1 has no API. It runs only in the browser at acemusic.ai; the interface is available in 14 languages.

acemusic.ai pricing, read on 11 November 2026, before tax.

Compared with Suno this is a strong offer: Suno Pro includes 20 downloads a month, while the cheapest paid ACE plan has unlimited downloads.

What else is in the app

• Two modes: Idea (just a description, the model writes the lyrics) and Custom (your lyrics, style, length from 30 seconds to 10 minutes).
• ACE Music v1.1 Preview: according to ACE it “plans every note before generating the sound”. It switches on automatically for covers and sheet music.
• Cover a song, Make a mashup, Sing sheet music (MusicXML or MIDI), singing to an image or video, recording your own idea.
• ACE-Step v1.5 XL: the open model, free in the app, with MP3 downloads even on the Free plan.

I tried sheet music with my own MIDI file of a simple folk melody. The model kept the melody in both variants, but in one it sang the words unclearly and in the other it replaced them with “la-la-la”.

Kinetyra

The link to ElevenLabs is a partner link: if you buy through it, QUANTSLANT earns a commission and you pay the same price. It has no influence on the content, including where the coverage is critical.