Business IEA Says Data Center Power Demand Could Double to 1,000 TWh by 2030, With AI Supplying Three Quarters of the Growth Chatbots ChatGPT Now Transcribes as You Talk: OpenAI Removes the 10-Minute Cap and Streams Words Live on Every Plan Business Ex-Alibaba Executive's Everful Raises $48.5 Million to Put AI Agents in Charge of Sourcing From 30,000 Factories AI Agents Google Built a Mirror Dimension of the Internet to Train Its Chrome AI Agents Without Breaking the Real Web Productivity Canva Finally Adds an Off Switch for Its Generative AI Features, and a Filter to Hide AI-Made Templates Business Anthropic Begins Trading Today: 980 Million Shares at $83, a $1.5 Trillion Valuation and the Largest IPO on Record News Google Books Blue Origin New Glenn to Launch TPU-Carrying AI Satellites in Early 2029 Security Hackers Used the AI Penetration Tool ARTEX to Disrupt European Commission Systems for Over Ten Days Business IEA Says Data Center Power Demand Could Double to 1,000 TWh by 2030, With AI Supplying Three Quarters of the Growth Chatbots ChatGPT Now Transcribes as You Talk: OpenAI Removes the 10-Minute Cap and Streams Words Live on Every Plan Business Ex-Alibaba Executive's Everful Raises $48.5 Million to Put AI Agents in Charge of Sourcing From 30,000 Factories AI Agents Google Built a Mirror Dimension of the Internet to Train Its Chrome AI Agents Without Breaking the Real Web Productivity Canva Finally Adds an Off Switch for Its Generative AI Features, and a Filter to Hide AI-Made Templates Business Anthropic Begins Trading Today: 980 Million Shares at $83, a $1.5 Trillion Valuation and the Largest IPO on Record News Google Books Blue Origin New Glenn to Launch TPU-Carrying AI Satellites in Early 2029 Security Hackers Used the AI Penetration Tool ARTEX to Disrupt European Commission Systems for Over Ten Days

ACE Music v1 Debuts Above Suno v6 on the Artificial Analysis Vocal Board at 1160 Elo, With a 16-Point Edge That Sits Inside the Error Bars

ACE Music v1 Debuts Above Suno v6 on the Artificial Analysis Vocal Board at 1160 Elo, With a 16-Point Edge That Sits Inside the Error Bars

ACE Studio's ACE Music v1 has taken first place on Artificial Analysis' blind-tested vocal leaderboard with an Elo of 1160, ahead of Suno v6 at 1144 and Suno v6-mini at 1132. The 16-point gap falls inside overlapping 95% confidence intervals, and the model is absent from the instrumental board where Suno v6 still leads. The Beijing team behind it began building its music foundation model only in March 2025.

For the first time since Suno's flagship arrived on the scene, a rival model has climbed above it on Artificial Analysis' vocal music leaderboard — and it comes from a Chinese startup most people had never heard of. ACE Music v1, built by the team behind the ACE Studio AI music workstation, debuted at the top of the AA-Music-Vocal v1.1 board with an Elo score of 1160 (±21) from 2,119 blind-comparison samples, edging out Suno v6 at 1144 (±18) and Suno v6-mini at 1132, according to a live check of the board conducted by Pandaily on November 16 and coverage from Chinese outlet QbitAI.

The caveat matters as much as the headline. Artificial Analysis builds its board from blind A/B tests: evaluators hear two tracks generated from the same prompt without knowing which model made each, and their preferences are aggregated into Elo-style scores. The 16-point gap between ACE Music v1 and Suno v6 sits inside overlapping 95% confidence intervals — roughly a 52-to-48 head-to-head edge — and the ranking service itself lists ACE Music v1 in a rank range of 1–2 and Suno v6 in a range of 1–3. ACE's sample count is also smaller than the 3,000-plus comparisons behind each of the other leading models. Statistically, the top of the board is a tie; journalistically, it is the first time anything has been placed above Suno v6 on that board since the model's September 9 release.

The rest of the vocal top six now reads: Suno v6-mini at 1132, Kunlun Tech's Mureka V9.5 at 1101, Mureka V9 at 1063, and Google's Lyria 3.5 at 1050. Notably, ACE Music v1 does not appear at all on the separate instrumental leaderboard, where Suno v6 still leads at 1148 — its claim is specifically about vocals.

Under the hood, ACE Music v1 takes a different route from most text-to-music systems. Instead of mapping a prompt straight to audio, it uses what the company calls Symbolic CoT: the model first generates an editable musical score — notes, beats and chords as symbols a language model can reason over and revise — and then renders audio from that score. Users can embed performance cues such as "hum softly" in parentheses, export songs to MIDI, and split them into stems. A v1.1 Preview goes further, planning music in ABC notation before generating audio so users can edit the score and regenerate. The company says it plans to open-source a lighter ACE Music v1 Flash model in late 2026.

The team's speed is part of the story. Beijing-based operator Timedomain Inc., founded in 2018 and led by CEO Guo Jing, only began developing its music generation foundation model in March 2025 — about 18 months before this debut — and Chinese media reports put its total funding at nearly 40 million Dollar. CTO Zhao Wenxiao came from Tencent's game engine work, while CPO Conger Sheng was one of the youngest signed songwriters at Warner Music China. The company also maintains the open-source ACE-Step music models.

Commercially, the ACE Music app gives free users 20 credits a day, enough for roughly 10 songs, with a Creator plan at 9.90 Dollar per month; songs made on paid plans carry commercial use rights, according to QbitAI. That pricing puts it well below the cost of most production music workflows, though the leaderboard result — a statistical tie with Suno on vocals — suggests the practical takeaway is competition, not a changing of the guard.

Either way, the blind-test result signals that the gap between the incumbent and the challengers in AI music has narrowed to something evaluators can barely distinguish. For a field where Suno has been the default reference point since 2023, the arrival of a credible #1-ranked challenger from an 18-month-old project is the real news — the error bars just mean everyone should read the ranking the way Artificial Analysis does: as a range, not a verdict.

Comments (0)

Log in to join the discussion

Log In

No comments yet