ACE Music v1 Overtakes Suno v6 on Vocal Leaderboard in Rising AI Audio Race

Artificial intelligence has made staggering leaps in generating complete musical compositions from simple text prompts, capturing the attention of both casual listeners and professional creators. Now, a relatively unfamiliar competitor has managed to edge out Suno, historically one of the biggest and most recognized names in the AI music generation space, at its own game on a prominent industry benchmark.

ACE Music v1, developed by ACE Studio, has officially claimed the top position on Artificial Analysis’ vocal music leaderboard. The model secured 1,160 Elo points, managing to surpass Suno v6, which currently sits at 1,144 Elo points.

The evaluation framework behind these rankings, known as the AA-Music-Vocal v1.1 benchmark, measures the overall quality of AI-generated songs that contain vocals. In these evaluations, human listeners are presented with tracks produced from identical prompts without knowing which specific underlying model created them. Evaluators then choose their favorite track based purely on listening preferences. Those accumulated user preferences determine an Elo-style rating, where higher scores directly reflect stronger listener preference across repeated blind tests.

Despite capturing the top spot, the competitive margin between the two AI models remains relatively narrow. Statistical confidence intervals for the models overlap, indicating a tight race at the pinnacle of AI vocal synthesis. Furthermore, Suno continues to hold a strong position, retaining the top spot on the separate instrumental music leaderboard provided by the same analysis platform.

ACE Studio, the company behind the rising challenger, was originally founded in 2018. For years, the firm concentrated its efforts on synthesizing realistic singing voices derived from user-inputted MIDI notes and lyrics. With the rollout of its latest flagship offering, the ACE Music v1 model, the company has expanded its capabilities significantly, enabling the generation of complete, finished songs from written descriptions or user-provided lyrics.

Putting the Challenger to the Test

In hands-on testing conducted by industry publications, ACE Music consistently produced more melodic and coherent songs tailored toward mainstream musical styles.

The vocal performances generated by the model generally sounded more natural to the ear, exhibiting far fewer of the synthetic artifacts that frequently plague AI-generated singing. In older or competing models, synthetic artifacts often manifest as vocalists sounding as though they are struggling to pronounce their own lyrics, relying excessively on heavy autotuning, or outright hallucinating random words simply to fill phonetic spaces within the rhythm.

Furthermore, ACE demonstrated a superior capability for handling song endings cleanly. While platforms like Suno occasionally introduce strange ambient noises, unnecessary lyrical repetitions, abrupt musical shifts, or unexpected hallucinations near the finish line of a track, ACE generally maintained its structural integrity and delivered much cleaner conclusions.

Industry experts note that crafting a convincing, catchy chorus is relatively easy for modern generative models compared to the much harder task of maintaining a cohesive musical idea throughout an entire composition, especially after the lyrics have concluded and the arrangement must wind down naturally.

However, the model is not without its limitations. Suno maintained a distinct advantage when testers pushed the models toward less conventional, heavier genres.

During melodic death metal tests, ACE consistently produced tracks featuring a woman singing clearly in a pleasant, clean voice, completely missing the aggressive, distorted vocal delivery expected from the requested style. Across several trials, the model frequently confused heavy metal subgenres. By contrast, Suno offered more creative arrangements in proper heavy styles, maintaining better stylistic consistency across similarly demanding, aggressive prompts.

These comparative evaluations highlight the reality that subjective performance varies widely by genre, proving that a single leaderboard victory does not automatically make one AI model universally superior for every musician or producer.

More Than a Simple Song Generator

Beyond raw vocal quality, ACE Studio’s most intriguing innovation may lie within its broader songwriting workflow rather than its prompt-to-audio output alone.

The platform features an "Idea" tool capable of developing a basic conceptual premise into fully realized lyrics and a structured composition. This feature intelligently identifies individual song sections—such as verses, choruses, and bridges—and details how they should be arranged structurally. Users can then iteratively refine the material rather than being forced to repeatedly generate completely new, randomized tracks from scratch.

The newer v1.1 Preview version pushes this workflow even further by establishing a formal musical score before generating the final audio file.

Developers explaining the system note that each generation acts as a fresh performance, meaning the exact sonic texture may differ slightly even when the underlying musical score remains entirely identical. This approach allows users to actively edit melodies and chords, regenerate individual performances, export raw MIDI files, and separate vocal stems from instrumental tracks for subsequent post-production work in traditional digital audio workstations.

This advanced workflow positions ACE closer to an intelligent music production assistant rather than a simple prompt-to-song novelty application.

Open-Source Ambitions and Future Releases

While the flagship ACE Music v1 model remains closed-source and lacks a public API—meaning users are restricted to generating tracks directly through the official website—its developers are simultaneously pursuing a parallel open-source strategy.

ACE Studio has partnered with StepFun to release ACE-Step 1.5, a separate family of music models whose underlying code and weights are publicly available to the community. Unlike the proprietary flagship system, these open-source iterations can be run locally on users’ own hardware, though they generally operate on a smaller, weaker model scale compared to the commercial version.

The company has also committed to expanding its open-source footprint further. Developers have stated plans to open-source "ACE Music v1 Flash" in late 2026, describing a specialized version engineered to strike a deliberate balance between high audio quality, rapid generation speeds, and manageable local hardware requirements.

ACE Music is currently accessible through its browser-based application, which offers a limited free tier alongside paid subscription plans. Commercial usage rights for songs generated via the flagship model require an active paid subscription, while the ACE-Step initiative provides a separate free generation path for open-source enthusiasts.

With ACE Music v1 currently holding a 16-point Elo lead over Suno v6 in vocal generation benchmarks, the competition in generative audio continues to heat up. The promised arrival of the Flash release later in the year will give local users and developers another powerful model to evaluate as the race for AI music supremacy moves forward.