Suno v6
A music AI was
rebuilt from its
training data up.
Suno's new v6 family was trained under licensing deals with Warner Music Group, BMG and Believe. The striking part comes next: the older models are being retired. A rebuild born out of settlement terms — and what precedent it sets for the copyright fights around generative AI.
The settlement asked for
a rebuilt model
On September 8, 2026, Suno released a new family of music models — v6, v6-wild and v6-mini. According to the official announcement, all three were trained under partnerships with Warner Music Group, BMG and Believe. CEO Mikey Shulman describes them as "faster, more expressive and higher quality," pointing to plain-language editing commands and the ability to generate songs from images and video.
But the substance here is not audio quality. The existing models are being retired — meaning the sonic character users have built around goes with them. As TechCrunch reports, that was part of the commitment Suno made when Warner settled its copyright suit in November 2025: ship more advanced licensed models, and retire the old ones. It is one of the few cases where a generative-AI copyright dispute has actually forced a company to retrain.
The split of roles is explicit. v6 is the flagship for Pro and Premier subscribers, reliable and polished across genres. v6-wild, also for those tiers, is deliberately less predictable and more exploratory. v6-mini is the lightweight model available on the free tier.
The scale of the swap
"Retiring a model" is not a mild phrase for the people using it. In generative AI, an older version usually lingers as an option after a new one ships. Here, the disappearance of the old version is itself part of the deal. Send the same prompt and you will not get last year's song back — that is what settlement compliance looks like in practice.
A portion of Suno's revenue goes to the partners, and those rightsholders decide how to pass it on to artists. Of the three, Believe sits closest to artists: those working with it can choose whether to take part in the new Suno experiences — which is what separates this from a plain wholesale of catalog.
| Earlier models | The v6 family |
|---|---|
| Training-data provenance was the subject of litigation | Trained on licensed catalogs from three partners |
| No revenue flowing back to rightsholders | A share of revenue goes to the partners |
| One line serving every user | Split into flagship, exploratory and free |
| Assumed to stay available | Being retired and replaced in phases |
What the lawsuit demanded was not an apology or a payout,
but a retrained model.
How to read the precedent
Why now
Copyright suits over generative AI have mostly ended in cash settlements or a service shutting down. This is a third outcome: rebuild and replace. Ship a model trained on swapped-out data, retire the old one. If that shape sticks, rightsholders in images, video and text may start asking for the same landing.
Who it helps, and how
Companies using Suno commercially — ad soundtracks, in-store music — get less rights uncertainty. Individual creators whose work leaned on the character of a specific older model lose that sound when it retires. Free-tier users move to v6-mini more or less automatically, and will notice it first as "the music sounds a bit different."
What to do next
If a live campaign uses Suno tracks, test whether the v6 line can reproduce the same texture before the swap reaches you, and keep local copies of what the old models made. Watch the UMG and Sony litigation too: its outcome bears directly on which catalogs are available to train on later, which is a real input to renewal decisions.
The counterpoint
None of this settles the copyright question. Universal Music Group and Sony Music remain in litigation with Suno, splitting from Warner and BMG, who settled. And the claim that no Universal or Sony music went into training the v6 suite is Suno's own account; no public means of independent verification has been offered. Until training provenance can be checked from outside, statements like this stay a matter of trust.