Skywork AI's Mureka V9 Closes Within 1 Elo Point of Suno's Top Spot
Mureka V9 debuts at #2 on both Artificial Analysis leaderboards, sitting within 1 Elo point of Suno V5.5 for instrumentals — and its DAW-native workflow sets it apart from every rival.
- Mureka V9 debuts at #2 on both Artificial Analysis Music Leaderboards (Instrumental and Vocals), within 1 Elo point of leader Suno V5.5 on instrumentals.
- MusiCoT architecture plans full song structure before generating audio, improving coherence across longer tracks compared to token-by-token generation.
- DAW-native workflow is Mureka's biggest differentiator: MIDI export, up to 12-stem splitting, and an in-DAW Ableton agent no competitor offers.
- Multilingual edge: V9 supports 10+ languages including Chinese, Japanese, Korean, and Spanish, outperforming Suno on non-English music.
- Active copyright lawsuit: Attack the Sound LLC v. Kunlun Tech (filed Dec 2025) alleges training on copyrighted recordings; no liability found yet.
- Pricing: Pro tier at $9/month; free tier with ~50 daily credits. Available via Artificial Analysis leaderboard and the Mureka app/API.
Artificial Analysis just dropped its blind-vote leaderboard results for Mureka V9, the latest flagship music model from Skywork AI, and the numbers are striking. On both the Instrumental and Vocals leaderboards, V9 debuts at #2 , directly behind Suno V5.5, the reigning benchmark. In the Instrumental category, the gap is essentially a rounding error: just 1 Elo point separates the two models.
A Challenger Closes the Gap
The Artificial Analysis Music Arena uses an Elo rating system , the same approach used to rank chess players , where real users listen to pairs of AI-generated tracks from the same prompt and vote for the one they prefer, without knowing which model made it. Models are ranked using Elo ratings derived from blind user votes, where listeners choose their preferred track without knowing which model created it. It's one of the most honest benchmarks in the space because there's no self-reported metric to game.
The current standings tell a clear story:
- Instrumental: Suno V5.5 (Elo 1194), Mureka V8 (Elo 1167), Suno V5 (Elo 1156), Lyria 3 Pro (Elo 1123), Suno V4.5 (Elo 1087) , with Mureka V9 now slotting in at #2, displacing V8.
- Vocals: Suno V5.5 (Elo 1156), Mureka V8 (Elo 1144), Lyria 3 Pro (Elo 1089), Suno V5 (Elo 1086), Suno V4.5 (Elo 1077) , again with V9 now at #2.
Suno has held the top spot for a while, and its V5.5 model is a formidable target. Suno V5.5 brought longer single-pass tracks with no stitching artifacts on multi-minute generations, cleaner stems on the Studio editor, and improved automatic mixing so raw output needs less mastering. Mureka V9 hasn't dethroned it , but it's now close enough that the difference may come down to personal taste rather than raw quality.
The Architecture Behind the Sound
What makes Mureka technically interesting is MusiCoT , short for Music Chain-of-Thought. The idea borrows from reasoning models in the LLM world: instead of generating audio in one pass, the model first plans the song's structure before producing any audio. Unlike traditional autoregressive models that generate audio step-by-step, MusiCoT pre-generates the overall music structure before finalizing the audio tokens for decoding, which significantly improves structural coherence and instrumental arrangement.