Suno’s v6 licensed music rollout arrived September 9 with three new AI music models and the company’s first formal product launch alongside record-industry partners. Suno says the models were rebuilt on a new dataset that includes material licensed from Warner Music Group, BMG and Believe, plus user data. That is a meaningful change for a company whose earlier systems have drawn copyright claims, though it does not answer every question about what went into the new training set.
Jack Brody, a Suno executive, told The Verge and Engadget that v6 was trained from the ground up with data different from that used for prior models. He identified licensed partner material and user data, but did not provide a complete accounting of the dataset or technical details. The Verge reported that it remains unclear whether the v6 data is entirely free of improperly obtained material.
That distinction is doing real work. The Hollywood Reporter characterized the models as trained exclusively on licensed music, while Suno’s own representatives described licensed content and user data as components of a newly assembled dataset. The public reporting does not establish the makeup of every item used to train v6.
What is the difference between Suno v6, v6-wild and v6-mini?
The standard v6 model is Suno’s flagship offering. V6-wild is intended to produce less predictable, more experimental results. V6-mini is the faster, resource-light option for free accounts; Engadget reported that the standard and wild models require Pro or Premier subscriptions.
All three models are part of the same September rollout, and Suno plans to retire its older models. In limited hands-on testing, The Verge found the new system better at hitting the surface markers of requested genres, but reported that it still failed to create deliberately off-key, dissonant or rhythmically loose music when asked. The reviewer also heard harsh artifacts in vocals. That is one reviewer’s test, not a benchmark.
What can Suno v6 do differently?
Suno says users can edit a portion of a generated song through a plain-language chat rather than recreating the entire track. The tools can replace a single lyric or alter an individual element such as a guitar part. Users can also combine pieces from separate Suno creations into a new result and generate music from text, audio, images or video, according to Billboard, The Verge and Engadget.
The commercial arrangement extends beyond training. Suno chief product officer Jack Dolby told Billboard that its music-industry partners will receive revenue sharing, but said those payments are tied to the broader partnership rather than payment for training data itself. Suno says it intends to build authorized remixing, sampling and derivative-work tools using music from participating partners. The financial terms and the eventual product designs have not been disclosed.
The partnerships arrive after years of conflict. Engadget reported that major labels sued Suno and rival Udio in 2024 over alleged copyright infringement involving training data. Warner dropped its case against Suno in 2025 as part of a licensing agreement, according to Engadget. Suno’s v6 release gives the company a label-backed path forward, but it does not settle the wider arguments around its earlier models or disclose the full provenance of the new model’s data.
This story draws on original reporting from The Verge.
