Artificial Analysis AI Music Leaderboard: Suno v6 Leads
Artificial Analysis released AA-Music v1.1. Suno v6 leads both vocal and instrumental categories, with Mureka V9.5 right behind. Here is the full breakdown.
👁 116 reads
Artificial Analysis released an updated AI music generator leaderboard on October 7, 2026. Suno v6 takes the top spot in both categories: vocal tracks (1142 Elo) and instrumental music (1140 Elo). Following closely are Suno v6-mini (1116) and Mureka V9.5 (1103), which jumped by 51 points compared to its previous version. I looked closely at what these benchmark numbers mean in practice, especially for non-English lyrics, and ran my own tests.
What Artificial Analysis announced
Artificial Analysis is an independent platform that has evaluated AI models across text, image, video, and audio. Its music leaderboard works like an arena: listeners hear two anonymized audio clips generated from the exact same prompt and vote for the better track. Thousands of blind pairwise comparisons form an Elo rating scale, much like the one used in chess.
The updated release features two separate leaderboards: AA-Music-Vocal v1.1 (songs with vocals) and AA-Music-Instrumental v1.1 (instrumental only). Notable changes over the initial version include:
- 1,000 new, more challenging prompts spanning 17 genres, from pop and hip-hop to classical and jazz.
- Evaluations restricted entirely to a recruited listener panel, accounting for around 37,000 blind comparisons (16,000 for vocal, 21,000 for instrumental). Votes from the public arena no longer influence the official score, reducing brigading risks.
- Over 2,000 head-to-head comparisons per model, narrowing the error margin to roughly ±15 Elo points.
- MiniMax Music 3.0 serves as the anchor point set at an Elo of 1000. Tracks are approximately 3 minutes long, and for vocal generations, the model writes its own lyrics.
These new scores cannot be compared directly against the previous release. In the September run, Suno V5.5 led with 1170 Elo ahead of Mureka V9 at 1159, but that iteration relied on a different anchor model and different prompts.
Results: the top tier in numbers
| Rank | Vocals (AA-Music-Vocal v1.1) | Elo | Instrumental (v1.1) | Elo |
|---|---|---|---|---|
| 1 | Suno v6 | 1142 | Suno v6 | 1140 |
| 2 | Suno v6-mini | 1116 | Suno v6-mini | 1109 |
| 3 | Mureka V9.5 | 1103 | Mureka V9 | 1081 |
| 4 | Mureka V9 | 1052 | Mureka V9.5 | 1080 |
| 5 | Lyria 3.5 (Google) | 1052 | Lyria 3 Pro | 1078 |
| 6 | Lyria 3 Pro | 1046 | Lyria 3.5 | 1076 |
| 7 | Eleven Music v2.5 | 1012 | Eleven Music | 1021 |
| 8 | Eleven Music | 1001 | Stable Audio 3 Large | 1017 |
Source: Vocal Leaderboard and Instrumental Leaderboard by Artificial Analysis, accessed October 8, 2026.
Key takeaways beyond the top spot:
- Mureka V9.5 is the only serious challenger to Suno right now. In vocals, it sits 13 points behind Suno v6-mini, which is within the ±15 margin of error - effectively a tie for 2nd place. In instrumental, V9.5 saw no notable improvement over V9.
- Ranks 3 through 6 in instrumental are a statistical tie. Mureka V9, V9.5, Lyria 3 Pro, and Lyria 3.5 all cluster between 1076 and 1081 Elo.
- MiniMax Music 3.0 is the top-ranked open-weights model, landing at 11th in vocals and 13th in instrumental. The open Stable Audio 3 Medium sits close behind (998 in instrumental).
- ElevenLabs features three model versions on the board, but none surpasses 1021 Elo.
- Udio is not listed in this benchmark run.
How the models handle non-English lyrics: our test
Public benchmarks like this have a clear blind spot: they focus almost entirely on English. Evaluators grade tracks where the model generates its own English lyrics. If you produce songs, podcast intros, or social videos in other languages, raw audio fidelity matters far less if the lyrics are slurred or mangled.
To test pronunciation accuracy, we tested it in Polish using an identical test song (2 verses and a chorus, 52 words) across every accessible model over two passes. Vocals were separated using Demucs, lyrics were transcribed via Whisper large-v3, and audio fidelity was scored by two independent AI judges. Here is the summary of the measured results (note: prompts and lyrics were originally tested in Polish):
- Eleven Music v2.5 - 103 out of 104 words correct, 5/5 pronunciation score from both judges. It delivered the clearest pronunciation, despite ranking 7th on the AA leaderboard.
- Lyria 3 Pro - 103 out of 104 words correct, sticking closest to the specified genre style.
- ACE-Step 1.5 (free, run locally on an RTX 4090) - 101 out of 104 words correct, rendering a full track in 8 seconds.
- MiniMax Music 3 (run locally) - 102 out of 104 words correct, though it strayed from the requested mood and consumed up to 20 GB of VRAM.
Detailed benchmark tables, audio samples, runtimes, and compute costs are detailed in my AI music generator ranking. Suno v6 does not provide a public API, so it was excluded from this programmatic pipeline - see my guide on generating non-English tracks in Suno for tips. Mureka was also not evaluated here, as its API requires account pre-funding and does not officially list Polish among its 10 supported languages.
Practical recommendations:
- If overall production polish and English lyrics are your top priority - Suno v6 currently leads the space, with Mureka V9.5 standing out as a capable alternative. See what changed in the latest version in our review of Suno v6, alongside pricing details in how much Suno costs.
- If vocal clarity in non-English tracks matters most - test Eleven Music (my test) or Lyria through Google Gemini (walkthrough). Both models transcribed nearly flawlessly in our tests.
- If you need quick instrumental backing tracks for short-form video - Stable Audio 3 running locally can generate a 60-second piece in 4 seconds (Stable Audio hub).
- If you want an entirely local, self-hosted pipeline - look at ACE-Step 1.5 or MiniMax Music 3 (licensing and setup guide).
Things to keep in mind
- Elo scores reflect subjective listener preference, not objective technical fidelity. Another independent benchmark, “A Thousand Listeners” from September 1, 2026, placed Mureka V9.5 ahead of Suno, though it matched it against the older Suno V5.5. Outcomes vary significantly depending on prompt style and listener demographics.
- Commercial rights remain a key concern. Suno faces copyright lawsuits filed by Universal and Sony on September 18, 2026, and free tiers across most services prohibit commercial exploitation. Review the legal context in our breakdown of Suno and copyright.
- This ranking reflects data as of October 2026. Given the rapid release cadence across generative audio, leaderboard positions can shift within weeks. I will keep updating my own benchmarks as new models drop.
For a complete overview of AI music platforms, features, and free tiers, read our comprehensive guide to AI music generators, or explore production workflows for podcasts in our piece on AI tools for music production.
Best AI Music Generators 2026: Suno v6, Mureka & Eleven MusicBenchmarks
ACE Music v1 tops the AI music leaderboard ahead of Suno v6: my testNews
Suno v6: Three Models, Label Deals, and What’s ChangedNewsThe 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.