1. ACE Music v1 takes the top spot on a vocal leaderboard
ACE Music v1, the full-song generator built by ACE Studio, has moved into first place on Artificial Analysis' vocal-music leaderboard, scoring 1,160 Elo points against 1,144 for Suno v6 as of November 11, 2026. The 16-point margin is narrow, but it marks the first time a lesser-known developer has displaced Suno at the top of that ranking.
Suno has for some time been the name most people associate with AI-generated songs, so a challenger overtaking it on a benchmark built around listener preference is notable. ACE Studio's model writes and performs complete tracks, vocals included, from either a plain-language description of the desired song or a set of lyrics supplied by the user.
The result does not mean the contest is settled. The two scores sit close together, the statistical confidence intervals around them overlap, and Suno continues to lead a separate leaderboard that ranks instrumental music. What the new ranking does show is that the gap between the market's best-known product and its rivals has effectively closed when it comes to songs with singing.
2. How does the Artificial Analysis music benchmark work?
The benchmark, labeled AA-Music-Vocal v1.1, asks human listeners to compare songs produced by different models from the same prompt without being told which model made which track. Each listener simply chooses the version they prefer, and those head-to-head picks are converted into an Elo-style rating.
Elo systems were originally designed to rank chess players, and they work the same way here: a model gains points when it wins a matchup and loses points when it is beaten, with larger swings for upsets against stronger opponents. Over many comparisons, the ratings settle into an ordering that reflects which systems listeners tend to favor.
Because the method depends on aggregated human taste rather than technical measurements, small gaps between scores should be read cautiously. A lead of 16 points with overlapping confidence intervals means listeners preferred ACE Music v1 somewhat more often, not that it wins every comparison or that every musician will hear it as clearly superior.
3. Who is ACE Studio?
ACE Studio was founded in 2018 and built its early reputation on a narrower problem than full song generation. Its original software synthesized singing voices from MIDI notes and lyrics, giving producers a way to create vocal lines without hiring or recording a singer.
That background in voice synthesis helps explain where ACE Music v1 appears strongest. Rather than starting from a general-purpose audio model and adding vocals later, the company spent years on the specific challenge of making artificial singing sound like a human performance. ACE Music v1 extends that work into complete compositions, handling melody, arrangement and vocals together.
For readers less familiar with music software, MIDI is a long-standing digital standard that records which notes are played, when and for how long, without containing any actual sound. A tool that turns MIDI notes and lyrics into a sung vocal is therefore doing the work of a performer: interpreting a written part and giving it pitch, timing, tone and expression. That is exactly the skill a full-song generator needs when it has to decide how a line should be sung.
4. Where ACE Music v1 beats Suno in hands-on testing
Early hands-on comparisons by reviewers broadly matched the leaderboard. Across mainstream styles, ACE Music v1 tended to produce more melodic, more coherent songs than Suno, and its vocals sounded more natural.
The difference was most audible in the small flaws that often give AI music away. ACE's singers showed fewer mispronounced words, less of the heavy autotune effect that can make generated vocals sound synthetic, and fewer stray filler sounds slipped in between lines. Those details matter because listeners tend to notice an unnatural vocal faster than any other weakness in a track.
Song structure was the second clear advantage. ACE Music v1 generally held a track together from start to finish and wrapped it up with a clean ending. Suno, by contrast, sometimes drifted in the final stretch, adding odd noises, repeating sections without reason, inventing material that did not fit, or changing direction abruptly just before the close.
That pattern points to a broader challenge in generative music. Producing a catchy chorus is comparatively easy for current models, since it is a short, self-contained burst of melody. Sustaining one musical idea across verses, bridges and an outro, and then landing the ending, requires the model to keep track of what it has already done over several minutes of audio, which is far harder.
5. Where Suno still has the edge
The picture shifted once testing moved away from popular genres. Suno handled unconventional styles more convincingly, and the clearest example came from a melodic death metal prompt.
In that test, ACE Music v1 delivered clean, pleasant female vocals, which is the opposite of the harsh, aggressive delivery the genre is built around. The model also tended to blur the lines between metal subgenres, producing tracks that did not clearly belong to the style requested. Suno's heavy arrangements were more inventive, and it stayed more consistent when given similar prompts several times in a row.
This kind of split is common among AI models trained on large music libraries. Systems usually perform best on the styles that are most heavily represented in their training material and struggle with smaller niches whose conventions differ sharply from the mainstream. For musicians working in those niches, an overall leaderboard win may say little about which tool will suit them.
The reviewers were also clear that their comparisons were subjective listening sessions rather than controlled benchmarks. A model topping a ranking does not make it the better choice for every musician or every kind of project.
6. Tools that make ACE Music v1 more than a prompt box
Beyond raw output quality, ACE Studio has positioned its product as a production assistant rather than a one-shot song generator. Several features push in that direction.
An Idea feature takes a basic concept and expands it into lyrics and a structured arrangement, labeling each section of the song so the user can see how it is laid out. Instead of regenerating an entire track every time something sounds wrong, users can refine individual parts of that material.
A preview release labeled v1.1 adds another step: it produces a musical score before generating any audio. The developers caution that the score does not lock in the final sound, noting that each generation is a new performance, so a track may sound different even when the underlying score stays the same.
Further editing tools let users adjust melodies and chords, regenerate the performance, export the music as MIDI for use in other software, and separate the vocals from the instrumental backing. Together, those options give producers more direct control than tools that only accept a text prompt and return a finished file.
This approach reflects how professional music is actually made. Producers rarely accept a first take as final; they rework a chord progression, rewrite a melody line or rebalance the mix until a track sounds right. A generator that only offers a full regeneration forces the user to gamble on a new result every time, while one that exposes the score, the separate stems and the MIDI data lets the user keep what works and change only what does not.
7. Is ACE Music v1 open source?
No. The flagship ACE Music v1 model is closed source, has no public API, and can only be used through ACE Studio's official website. Developers who want to build it into their own applications or run it on their own machines cannot do so.
There is an open alternative from the same team. ACE Studio and StepFun have released ACE-Step 1.5 with its code and model weights publicly available, which means it can run on local hardware. It is a weaker model than the flagship, but it offers a free route for users who prefer to keep their work on their own computers.
A middle option is on the way. The developers have said they plan to open-source ACE Music v1 Flash in late 2026. The Flash version is designed to balance output quality against generation speed and hardware requirements, making it a candidate for people who want something closer to the flagship without relying on the company's servers.
8. How much does ACE Music cost?
ACE Music v1 runs as a browser app with a limited free tier alongside paid subscriptions. Anyone planning to use songs from the flagship model commercially needs a paid plan, since commercial rights are tied to those subscriptions.
ACE-Step remains a separate, free option for users who are comfortable running an open model locally and can accept lower quality in exchange for avoiding subscription costs and keeping full control of their setup.
9. What comes next for AI music generation
For now, ACE Music v1 holds a 16-point lead over Suno v6 on the vocal leaderboard, a margin small enough that further updates from either company could reverse it. Suno's continued lead in instrumental rankings and its stronger handling of niche genres mean the competition between the two is likely to be fought style by style rather than decided by a single score.
Musicians weighing the two tools may find the choice depends less on rankings and more on their own work. Someone writing mainstream pop or singer-songwriter material may value ACE's cleaner vocals and tidier endings, while someone producing heavier or more experimental music may still get better results from Suno.
The next milestone to watch is the planned open-source release of ACE Music v1 Flash later in 2026. If it arrives on schedule, users who prefer to run models locally will have another, stronger option before the year ends, and the pressure on closed music generators to match both quality and flexibility will grow.