ACE Music v1 Beats Suno v6 on Vocal Leaderboard: 1160 vs 1144 Elo
The ACE Music v1 song generation model took first place on the Artificial Analysis vocal leaderboard (AA-Music-Vocal v1.1). It scored 1160 Elo points, while Suno v6 scored 1144. The model creates songs from a style description or from ready-made lyrics, and you can try it for free on the project's website. The gap is small, but if the weights are really released, closed services will face a strong open alternative.
AI-processed from TG: @bezsmuzi (Русский ИТ бизнес); edited by Hamidun News
Music generation has long been the territory of closed services: you go to a website, write a description, get a track, and know nothing about what's inside. Now this market has a new contender. The ACE Music v1 model took first place on the Artificial Analysis AA-Music-Vocal v1.1 vocal leaderboard, scoring 1160 Elo points. Suno v6 scored 1144 on the same leaderboard.
What ACE Music v1 showed
ACE Music v1 generates songs in two ways: from a text description (genre, mood, performance style) and from ready-made lyrics that it turns into a song with vocals. For creators of videos, games, or ads, these are the two most common scenarios. Either you have an idea of the sound, or the words are already written.
An important detail: the leaderboard is vocal. It doesn't assess music in general but how the singing sounds: naturalness of the voice, delivery, fit to the lyrics. Vocals are usually what gives away an AI-generated song fastest, so leading this category is a notable argument for the model.
You can try ACE Music for free on the project's website, which has a free tier. The author of the original post has not tested the model himself and only noted that he wants to try it, so the material contains no independent impressions from live use.
How large the gap is
The numbers should be read soberly. A 16-point Elo difference is small. Under the standard Elo formula, such a gap means that in a head-to-head comparison the leader wins about 52% of the time, that is, only slightly more often than it loses. Strictly speaking, this is not a rout but rather a near tie with a slight edge. Moreover, the leaderboard reflects a specific set of tests and the preferences of the evaluators, not absolute quality on every task.
The reasonable conclusion, therefore, is this: ACE Music v1 has at least entered the same quality class as Suno v6, and according to Artificial Analysis it is slightly ahead. For the market, this is already news, because closed services have had almost no rival of this kind in vocals until now.
Open weights: what is promised
The most interesting part for developers is the plans around openness. The creators promise to publish the weights within the next few weeks. Separately, the website says that a lighter version, ACE Music v1 Flash, will be released as open source at the end of 2026.
Note the wording: for now these are promises, not a release. The available data does not say what license the weights will carry, what graphics cards will be required, or whether the output can be used commercially. Until these terms are published, it is too early to build the model into a product.
What to do in practice
If you are building a product that needs music (games, videos, podcasts, creative apps), it makes sense to compare ACE Music and Suno on your own tasks right now: take 5–10 of your own prompts or lyrics and listen to the results blind. A 16-point Elo gap will not replace your own ear or your own requirements for style.
If it matters to you to run the model yourself, without depending on someone else's API or subscription, keep an eye out for the weights release. Only then will it be clear how closely the open version matches what the website shows.
For the market as a whole, the signal is simple: in music generation, as earlier in images and text, closed leaders are starting to be caught by models that promise to open up. If that promise is kept, the price and accessibility of AI music could change noticeably.
- Nov 21, 2026Open-Source MiniMind Lets You Train Your Own LLM from Scratch for $0.42 on an RTX 3090
- Nov 21, 2026TensorRT-LLM Beats vLLM on H100: 2780 vs 2400 Tokens/s on Llama 3.3 70B
- Nov 21, 20261C Developers Switch to AI Agents: McKinsey Sees Just a 3% Productivity Gain
- Nov 21, 2026Empatra Studio 0.29 Cuts Frontend Agent Costs by 59% and Runs Twice as Fast
Key Questions Answered
What ACE Music v1 showed?
ACE Music v1 generates songs in two ways: from a text description (genre, mood, performance style) and from ready-made lyrics that it turns into a song with vocals. For creators of videos, games, or ads, these are the two most common scenarios. Either you have an idea of the sound, or the words are already written. An important detail: the leaderboard is vocal. It doesn't assess music in general but how the singing sounds: naturalness of the voice, delivery, fit to the lyrics.
How large the gap is?
The numbers should be read soberly. A 16-point Elo difference is small. Under the standard Elo formula, such a gap means that in a head-to-head comparison the leader wins about 52% of the time, that is, only slightly more often than it loses. Strictly speaking, this is not a rout but rather a near tie with a slight edge. Moreover, the leaderboard reflects a specific set of tests and the preferences of the evaluators, not absolute quality on every task.
What are the key details on Open weights: what is promised?
The most interesting part for developers is the plans around openness. The creators promise to publish the weights within the next few weeks. Separately, the website says that a lighter version, ACE Music v1 Flash, will be released as open source at the end of 2026. Note the wording: for now these are promises, not a release. The available data does not say what license the weights will carry, what graphics cards will be required, or whether the output can be used commercially.
Test Frontier LLMs & Autonomous Agents in One Workspace
Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.