AI
Suno's Latest AI Music Model Comes With Major Record Labels on Board
The music platform debuts version 6, trained on licensed content from Warner Music Group and other industry partners—though questions remain about whether older training data was fully cleared.
The Verge
Suno has unveiled v6, a new AI music generation model developed with backing from the record industry. Unlike earlier iterations, the system was built using licensed training data from major music organizations, marking a shift toward more transparent development practices.
Industry Partnerships and Data#
The v6 model incorporates content licensed from Warner Music Group, BMG, and Believe, along with user-generated material. However, it remains unclear whether this represents a complete break from the company's past reliance on training sources of questionable legal standing. According to Suno's Jack Brody, the system was "trained from the ground up, with a new set of data that does not include the same data that our previous models were trained on."
Three Models at Different Tiers#
The new release introduces a three-tier architecture. The free v6-mini prioritizes speed and efficiency, generating simpler tracks with visible AI artifacts. v6-wild aims for unpredictability and experimental results. The standard v6 sits between them, offering a balance of quality and control.
A major improvement across all three is genre recognition. The models now handle specialized styles—from hyperpop to krautrock—with genuine understanding of their sonic conventions, a significant leap from earlier versions that struggled with niche categories.
Persistent Limitations#
Despite these advances, Suno's v6 continues to fall short in one critical area: generating the subtle, human irregularities that define natural music. The model cannot produce intentionally off-key vocals, detuned instruments, or rhythmic imperfection, no matter how explicitly prompted. Even requests for specific flaws—such as monotone vocals or dissonant piano—are routinely disregarded. The system excels at harmonic and rhythmic precision but at the cost of emotional authenticity.
Additionally, the model has actually become more prone to the harsh, unnatural vocal artifacts common in AI-generated content, a step backward from the previous version.
