--- title: "China's Mureka Outsang Suno on the Global Music AI Bench — But Can It Out-Earn It?" date: 2026-09-27 category: Generative Media site: NeuroAI canonical: https://neuroai.site/a/na-aigc-ai-music-china language: en --- # China's Mureka Outsang Suno on the Global Music AI Bench — But Can It Out-Earn It? > Kunlun Tech's Mureka V8 ranked #1 on Artificial Analysis's vocals and instrumental music leaderboards in March 2026, per the company's Q1 filing. The harder question is whether AI music can turn benchmark wins into recurring revenue. You type a feeling and a melody comes back. The singer breathes at the ends of lines. The mix sounds like a person made it. For years that bar belonged to American labs. This year a Chinese model (大模型) quietly took the top spot on the independent global benchmark. The name is Mureka, and the company behind it just posted numbers that suggest AI music is becoming a business, not a demo. ## Key takeaways - Kunlun Tech's **Mureka V8** ranked **#1 on both the vocals and instrumental leaderboards** of Artificial Analysis, an independent AI benchmark, in March 2026. - Per Kunlun's Q1 2026 filing, Mureka V8 outperformed international rivals including **Suno V5, Google Lyria 3 Pro and Minimax Music 2.5+**. - **Mureka V9** followed in March 2026, sharpening text control and mixing. - Kunlun's Q1 2026 revenue was **2.57 billion yuan (25.7亿元, ≈ US$362M / HK$2.8B)**, up **45.69%** year on year; overseas revenue reached **2.49 billion yuan (24.87亿元)**, up 49.29%. - Mureka's API is live; the company frames the product as an AI music platform, not just a generator. ## The benchmark that matters Artificial Analysis runs standardized, independently executed tests rather than accepting vendor-reported numbers. In March 2026, Kunlun said Mureka V8 reached #1 on its **vocals** and **instrumental** music leaderboards — the two dimensions that decide whether a generated track sounds like a product or a prototype. The company credited its in-house MusiCoT framework, which plans a song's structure before filling in detail. Whether the leaderboard still holds today is a moving target; the claim reflects the March 2026 snapshot reported in Kunlun's filing. ## From generator to platform Mureka is built by Kunlun Tech (昆仑万维), a company that began an AI pivot in 2020 and now runs an "AI-native entertainment" strategy spanning short drama (短剧), music, games and agents. Its short-drama and AI short-drama platform passed **US$570 million in annual recurring revenue** and more than **US$48 million in monthly revenue** in early 2026 — context that Mureka is not a lab experiment but part of a monetizing content machine. The music product is pitched two ways: a creation tool for users, and an API for businesses. Kunlun says Mureka's API went live in 2026, positioning it as a platform rather than a single model. ## Why "sounds human" is the hard problem Early AI music was easy to spot: thick arrangements, crowded melodies, no space between phrases. Mureka's stated direction is to copy how human musicians work — plan the whole structure, then refine — and to strip the tell-tale "AI sheen." The V9 update tightened arrangement logic and reduced over-stacked instrumentation. China's edge in this category is partly linguistic and cultural: training data and user taste for Mandarin lyrics and regional styles give domestic models a home-ground advantage that global rivals lack. ## The real competitor is the workflow, not the waveform Suno remains the best-known AI music product globally and has the largest user base. Mureka's bet is different: it is embedded in a parent company that already runs distribution — short drama (短剧), games, social — so a generated track can flow straight into a product rather than sitting in a folder. That vertical integration is harder to copy than a model checkpoint. This is also why the financials matter. A model that tops a benchmark but loses money is a research project. A model inside a company posting 45% revenue growth is a commercial claim. ## A quiet win, measured properly Mureka's rise is not proof that Chinese AI music has won — leaderboards move, and Suno still has the larger user base and the deeper ecosystem. What it does show is that the gap has closed on the one metric ordinary listeners care about: does it sound like a person made it? For the first time, a domestic model (大模型) can credibly claim the top independent score, and its parent company can pay for the effort out of real revenue rather than pure funding. ## What readers can do now - **If you make content, test Mureka's API against Suno for Mandarin or regional styles.** Local models often win on languages global tools under-serve. - **Treat AI music as a workflow, not a toy.** The companies pulling ahead are bundling generation, editing and licensing into platforms — evaluate the whole stack. - **Watch the copyright line.** Training-data rights for AI music are unresolved in China; confirm the platform's terms before commercial use. ## Honest limitations The leaderboard claim comes from Kunlun's Q1 2026 filing and associated financial-press coverage (Securities Times, China Financial Information Network); I did not independently re-run Artificial Analysis benchmarks, and live rankings change. Competitor version names (Suno V5, Lyria 3 Pro, Minimax Music 2.5+) are as stated in Kunlun's filing. Revenue and growth figures are from Kunlun's Q1 2026 report (announced 28 April 2026). The short-drama ARR and monthly-revenue figures are company-reported. Not investment advice. --- Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-aigc-ai-music-china Free to quote with attribution and a link to the original.