Suno v6 Is the First AI Music Model Built With Record Label Blessing, Here Is What Has Actually Changed

Key points

• Suno launched its v6 AI music model in mid-2025, trained partly on content licensed from Warner Music Group, BMG, and Believe.
• v6 splits into three tiers: v6-mini (free), standard v6, and v6-wild, each aimed at different use cases.
• Users can now edit individual parts of a song in plain language, without regenerating the whole track.
• v6 handles niche music genres far better than earlier Suno models, but still cannot produce genuinely off-key or dissonant results.
• Suno says it will retire all previous models once v6 is fully rolled out.

Suno has just released v6, its most significant update yet, and for the first time the company can say the model was built with actual permission from the music industry. Warner Music Group, BMG, and Believe all licensed content for the training data. That matters because Suno's earlier models faced serious questions about whether they were trained on copyrighted recordings without consent.

Suno executive Jack Brody told The Verge that v6 was "trained from the ground up" with a fresh dataset. Whether that means every single audio file is above board is still unclear. The company also used "user data," which it has not fully defined.

What are the three new models?

v6 is actually three separate tools in one release. Think of them as economy, standard, and experimental.

v6-mini, the free tier, produces results quickly but the audio sometimes has small glitches that give away its AI origins. v6-wild is supposed to sound a bit looser and less polished, the kind of thing a band might record in one take. In practice, the difference between wild and standard v6 is subtle enough that most listeners probably would not notice.

All three models handle music genres significantly better than before. Feed any of them a prompt asking for krautrock (a 1970s German experimental rock style) or hyperpop (a hyper-produced, glitchy pop sound), and v6 nails the key characteristics. Earlier Suno models often missed the mark entirely.

What does it still get wrong?

Perfection, oddly enough, is still the problem. Ask v6 for monotone vocals, no drums, or a piano that is deliberately out of tune, and it ignores you. Every track comes out harmonically clean and rhythmically tight, no matter how hard you push it toward messiness. Trying to get the model to make something genuinely flawed is like asking a spell-checker to leave in the typos.

The AI-generated vocal artefacts, those faint harsh or watery sounds that reveal a voice was synthesised by software rather than recorded in a booth, are actually slightly more noticeable in v6 than in the previous version, v5.

What new editing features are there?

This is where v6 gets practically useful. You can now tweak a single guitar line or change one lyric without remaking the entire song. Just type the change you want in a chat window, the way you would message a friend. Suno calls this its plain-language editor.

The model can also combine several tracks from your Suno library into one new piece, and it can generate music from an image, a video clip, or another piece of audio, not just a text description. That opens up things like humming a melody into your phone and asking Suno to build a full track around it.

Suno says v6 starts rolling out now, with older models being phased out over time.

Common questions

Is the free version worth trying?

Yes. v6-mini is free for everyone and handles genre prompts well enough for casual experimenting, though expect occasional audio glitches compared with the paid tiers.

Does the record label deal mean the music is safe to use commercially?

Not automatically. Licensing content for training is different from clearing the output for commercial release. Check Suno's terms of service before using any generated track in a business context.