Suno shipped v6 on September 9, 2026, and killed every model that came before it in the same breath. Three new models. Seven new creation features. A Custom Model builder sitting in beta at the top of the picker, and a licensing deal with Warner, BMG and Believe holding the whole thing up.

Then I ran the numbers on what it costs to actually release a song, and found the part nobody covered.

Suno v6 is a family of three AI music models released on September 9, 2026, trained from scratch on licensed catalogue from Warner Music Group, BMG and Believe. It replaces every earlier Suno model. Paid users get v6 and v6-wild, free users get v6-mini, and all three add section editing, mashups, sampling, multimodal prompting and natural-language lyric edits.

That's the consensus answer, and it's accurate. What it hides is that on a Pro plan, 96% of the credits you pay for can never legally become a released track. I'll show that math in full below, because it changes which plan you should be on.

I run Roo Knows, and I've been pricing every Suno change since the download caps were announced on August 10, 2026. The cost tables further down are mine, computed from Suno's own published credit allowances and download tiers. They aren't reprinted from anywhere. Nobody has published them.

Suno v6 splits into three models, and the split isn't about quality alone, but about how tightly each model is tuned to what the average listener enjoys.

The flagship Suno v6 is the precise one: give it detailed instructions and it follows them, which makes it the right default for anyone who already knows exactly what they want to hear before they start. Pro and Premier only.

v6-wild is the interesting release. It's deliberately less tuned toward average preference, which makes it wilder and less reliable by design. Suno's chief product officer Jack Brody described the underlying problem as the model overfitting to what people enjoy on average, which drags every output toward pop. v6-wild exists to break that pull. Also Pro and Premier.

v6-mini is free for everyone, faster, lighter, and positioned as better than any other free music model on the market. Early testing backs that framing more than it backs the flagship, because one Japanese reviewer publishing hands-on notes on launch day found v6-mini surprisingly strong for a free tier while reporting they couldn't clearly distinguish standard Suno v6 output from v5.5.

That same reviewer found roughly half of their v6-wild generations came out broken and unusable, which is one tester's sample rather than a benchmark. It matches what "less tuned toward preference" should produce, though, and it's the kind of detail the launch coverage skipped entirely.

Use Suno v6 with a brief. Pick v6-wild when you're stuck and want something you wouldn't have asked for, and pick v6-mini when you're testing an idea you may throw away, which, given the download economics below, is most of the time.

Seven creation features shipped with Suno v6. Here's each one, what it does, where it's likely to break, and whether it changes anything about the way you actually work.

Section editing lets you change one part of a finished song in plain language while everything else stays put, so if you ask for a different chorus you get a different chorus rather than a whole new song. This is the biggest workflow change in Suno v6, because it ends the old loop of regenerating an entire track to fix eight seconds of it.

Lyric refinement changes one word or one line without rebuilding the song around it, which is narrower than section editing and considerably more surgical.

Mashups combine elements from different songs in one request, with you describing how they should fit together. Vocal from one generation, drums from another. This only works across your own songs, which matters more than it sounds, because that's a rights boundary rather than a technical one.

Sampling and audio isolation lets you sample a specific moment, isolate an instrument out of it, and build a new beat around it in a single pass. The example Suno leads with is sampling a riff at a timestamp, isolating the guitar, and constructing a beat from it.

Multimodal input covers text, audio, images, video. Video-to-music is the genuinely new input here, and it's the feature most likely to get used for reasons Suno didn't intend, since it turns any clip into a scoring brief.

Vibe-based creation lets you prompt from a feeling, a genre, or a mix of reference points rather than a structured spec. It counterweights the flagship's precision.

Better prompt comprehension is the seventh. Suno says the models read vocals, instrumentation, structure, mood and reference points more accurately, which is a soft claim, hard to verify, and the one place where the launch messaging outruns anything a user can check on day one.

This is the feature that went almost entirely unreported, and it's sitting at the top of the Suno v6 model dropdown tagged Beta.

Create Custom Model builds a private model trained on tracks you upload, costs 100 credits, needs a minimum of six songs, and takes roughly two to five minutes to finish training. Pro and Premier subscribers can hold up to three models at once, and bulk upload means you're not feeding the thing one file at a time. The finished model shows up as its own selectable entry in the picker alongside Suno v6 and v6-wild.

You must own what you upload. Custom Models are private and can't be shared between accounts, which closes the obvious abuse route of training a model on someone else's catalogue and passing it around.

Suno's own help documentation still describes this as a v5.5 feature, yet it's now live in the v6 picker, which means the thing carried straight across the model transition. What Suno hasn't confirmed is whether a Custom Model created today fine-tunes the v6 base or still sits on v5.5 underneath. That's the most useful unanswered question about Suno v6 right now, and no press coverage asks it.

Price it in songs rather than credits and it gets clearer. At roughly five credits a generation, the 100-credit training fee is 20 generations forgone, and filling all three model slots costs 300 credits, which is 12% of a Pro subscriber's entire monthly allowance against 3% of a Premier's. Cheap on Premier. Painful on Pro.

Neither of these appears in Suno's launch post, and neither made the press coverage. Both show up in the interface.

Max Mode is a quality toggle that consumes double credits per generation, so at the standard rate of about five credits per song it takes a single song to roughly ten. On Pro that cuts your monthly draft ceiling from 500 songs to 250.

Variety rewrites your style prompt automatically, and the launch-day tester who documented it recommends turning the setting off on the grounds that the rewritten prompts come back short and weaker than whatever you wrote yourself. If you've spent time building a prompt library, that's the setting that will quietly undo it.

I'm flagging both as reported rather than confirmed, since they come from hands-on documentation rather than Suno's own materials, and Suno hasn't described either one publicly. Check your own interface first.

There's a reason Max Mode deserves more attention than it's getting. Every Suno v6 plan sells you credits, and Max Mode is the only setting that changes what a credit buys. Double the spend per generation and your Pro draft ceiling halves from 500 songs to 250, which sounds bad until you remember you can only release 20 of them anyway. The credits weren't the binding constraint. They never were.

So Max Mode is close to free in practice for most subscribers. You're spending a resource you have far too much of to improve the small number of tracks that will actually leave the platform. If a song is a candidate for release, run it in Max Mode. If it's a throwaway idea, don't. That single rule probably extracts more value from a Suno v6 subscription than any prompt trick.

Variety is the opposite case, a setting that costs nothing and can quietly damage your output. Test it on a prompt you know well before trusting it on anything else.

Here's the part the launch coverage missed, and it's the reason to read past the feature list.

Suno v6 arrived after two changes that reshaped the economics: download caps that started September 3, 2026, and paid downloads at $2.99 each. Generating is cheap. Releasing isn't. Two separate budgets. The gap is enormous.

The published inputs are simple. Pro is $10 a month, or $8 billed annually, with 2,500 credits and 20 downloads a month. Premier is $30 a month, or $24 annually, with 10,000 credits and 60 downloads. Free gets 50 credits a day and seven downloads total for the lifetime of the account, with no commercial rights. A song runs about five credits. Extra downloads are $2.99 with no volume discount, a detail I verified when the purchase screen went live on September 3 and found the 5-pack and 10-pack priced at exactly the single rate multiplied out.

Nobody has published what those inputs mean together. So I did the math.

A Pro subscriber can make 500 songs a month and release 20 of them. Premier is worse on ratio, not better: 2,000 generations against 60 releases.

Turn Max Mode on and the ratio improves, because you're burning credits twice as fast against a fixed download cap. Pro drops to 12.5 to 1, Premier to 16.7 to 1. That's a strange incentive. The expensive quality setting is the one that wastes less of what you're paying for.

Multiply the download cap by five credits and you get the credits that can ever turn into something you're allowed to release.

On Pro, 2,400 of your 2,500 monthly credits can only ever become drafts. On Premier, 9,700 of 10,000. The credit allowance is a drafting budget. The download cap is the real product limit, and anyone comparing Suno v6 plans by credit count is measuring the wrong thing entirely, because the number that decides what you can actually put into the world is the one printed in much smaller type.

I want to be fair about the counterargument, because there's a real one. Most people making music on Suno v6 aren't releasing 20 tracks a month, and for them the caps never bind. Suno also left Premier's Studio export unmetered, so 32-bit/48kHz multitracks and stems come out without spending a download, which is a genuine escape hatch for anyone finishing tracks in a DAW. The caps are aimed at bulk streaming upload, not at hobbyists. That's a defensible design.

It's still worth knowing before you pick a tier, and the credit number on the pricing page tells you almost nothing about it.

Because overage is flat-rate, the tiers cross at a specific point.

Premier becomes the cheaper plan at 27 releases a month. Below that, Pro wins. Above it, Pro gets punishing fast. At 60 releases a month, staying on Pro costs $129.60 against Premier's $30.00, a $99.60 penalty for being on the wrong plan.

Per released song the picture is blunt. Pro at its cap is $0.50 a song. Push Pro to 100 releases and you're at $2.49 a song, which is within thirty cents of just buying downloads outright with no subscription at all.

Put out one track a week, 52 a year, and Pro costs $120 a year, working out to $2.31 per released song. Premier costs $360 and works out to $6.92 per released song for the exact same output. Premier only makes sense above 27 releases a month. Below that you're paying for headroom you'll never reach.

Studio 2.0 landed on August 13, 2026, roughly four weeks before Suno v6, and the two are designed to work together even though they were announced separately. It brought MIDI support, stem separation and audio effects into Suno itself, turning what had been a generator into something closer to a workstation.

Here's the part that matters for your bill. Premier subscribers can export 32-bit/48kHz multitracks and stems out of Studio without spending a download. No cap. No fee. For anyone who finishes tracks in Ableton, Logic or Reaper rather than shipping Suno's mixdown straight to Spotify, that's the only unmetered route out of the platform, and it quietly makes Premier a completely different value proposition than the download math alone suggests.

Think about what that means next to the caps. Suno metered the thing that goes directly to streaming services and left the thing that goes to a producer's DAW wide open. That's a deliberate split, and it tells you who Suno thinks the caps are aimed at. Bulk uploaders. Not musicians.

The natural-language editing in Suno v6 sits on top of these same controls rather than replacing them. Brody was candid about why. Telling a model to clip something at the third downbeat by 4.3 is painful in words, and dragging a waveform takes a second. So the two control schemes coexist: prompt for the creative decisions, drag for the surgical ones. Anyone expecting Suno v6 to make the DAW controls obsolete misread the release.

If you're on Premier and you've never opened Studio, that's the first thing to fix this week. You're paying for it.

Suno v6 was late. Badly late.

Warner settled with Suno on November 25, 2025, and buried inside that settlement was a commitment with a deadline attached, which was to launch new licensed models and retire the old unlicensed ones rather than simply carrying on as before with a signature on file. By March 23, 2026, 114 days had passed with no new model and v5 still running exactly as before. Music Business Worldwide's Tim Ingham read that gap as a litigation strategy rather than a product roadmap, arguing the deal bought Suno the appearance of industry peace without requiring it to actually ship anything.

Community sentiment through that window was split and impatient. Some Suno users on Reddit predicted a March or April arrival that never came. Others worried a licensed model would arrive creatively neutered, sanded down by rightsholder constraints into something safer and duller than what they already had. A few said outright they'd cancel the day it shipped.

That context changes how you read launch day. Suno v6 didn't arrive as a routine version bump. It arrived nine and a half months after the commitment that required it, three days after the download caps took effect, one day after the Believe partnership, and seven weeks after a Munich court ruled against Suno on six GEMA compositions. The sequencing isn't coincidence.

Which also explains the "trained from scratch" language Suno keeps repeating. It isn't a technical brag. It's the claim that the new models share no lineage with the ones Universal and Sony are suing over, and retiring every old model on the same day is what makes that claim clean. You can't accidentally serve an unlicensed generation from a model that no longer exists.

Every Suno v6 output carries audio watermarking and fingerprinting. Suno describes the technology as durable, tamper-resistant and inaudible. The partners are Audible Magic and Musixmatch, and Musixmatch's Sentinel screens both prompts and outputs against existing compositions, which lets distributors flag AI-generated material at intake rather than after release.

The payoff is distribution. Believe and TuneCore refused to distribute Suno music earlier in 2026. After the partnership announced on September 8, tracks made with Suno's industry-partner model became eligible for distribution through Believe and TuneCore, which reaches Spotify, Apple Music, Amazon Music and YouTube.

Connect those two facts and the download cap stops looking like a pricing decision. Watermarking makes every Suno v6 track identifiable off-platform. Download caps meter how many tracks can leave. Distribution eligibility gives them somewhere to go. That's not three separate announcements. It's one system, an identified and metered and licensed pipe running from a prompt box all the way to Spotify, engineered in that shape specifically so the rightsholders who spent 2025 suing Suno could put their names to it without losing control of the outflow. The $2.99 download isn't a paywall on a file. It's a toll on commercial release, and the labels needed the meter before they'd open the pipe.

Suno v6 was trained from scratch. That phrase is doing legal work, not just technical work.

BMG became the first major rightsholder to license Suno after the Warner settlement. BMG wasn't an RIAA plaintiff, so that deal covered prior use rather than settling live litigation.

What's absent matters as much as what's present. Suno v6 excludes Universal Music Group and Sony Music Entertainment recordings, and both remain plaintiffs against Suno in Boston with no trial date set. Fact discovery and depositions close September 30, 2026. Dispositive motions are due April 9, 2027. Suno is still arguing fair use rather than settling.

Revenue sharing starts now. Suno passes a portion of revenue to its partners, and the rightsholders decide how to distribute it downstream. No percentage has been disclosed. Suno's terms also note that rights holders may collect revenue from third-party distribution of outputs, which is worth reading carefully if you plan to monetize.

The artist opt-in layer is phase two, not today. Artists will be able to license their likeness and catalogue for remix experiences, paid through their labels. Suno hasn't dated it.

Retiring every old model is a one-way door. If v5.5 suited how you worked, there's no rollback, no legacy toggle, and no archive, and the launch-day reports suggesting Suno v6 output sits close to v5.5 in raw quality make the whole migration feel less like an upgrade you chose and more like a legal reset you were handed with some genuinely good features bolted on top.

Natural language editing has a ceiling. Brody was direct about it: you can't easily specify a clip at the third downbeat by 4.3 in words, and dragging a waveform is faster. Studio 2.0's DAW controls exist because prompting can't do everything, and Suno v6 doesn't change that.

Imitation prompts are still blocked. You can ask for characteristics, moods and genres. You can't order a track that sounds exactly like a named living artist.

The free tier's seven downloads are a lifetime allowance, not monthly, with no commercial rights attached. v6-mini being good is somewhat beside the point if you can only ever take seven songs out of the building.

Mashups and sampling stop at your own catalogue. You can pull a vocal from one of your generations and drums from another, but you can't feed in someone else's record and cut it up, which is the feature a lot of producers actually wanted. That boundary exists because of the licensing deal, not because the model can't do it.

And the quality question stays open. Suno v6 is marketed as faster, more expressive and higher quality, yet the only independent hands-on account published on launch day couldn't reliably separate it from v5.5 by ear. One tester isn't evidence of much. It's still the only evidence there is on day one, and it sits awkwardly next to a launch built almost entirely on the claim that these models are better.

Udio settled. Suno didn't, and the split now runs through everything.

Universal settled with Udio in October 2025 and struck a deal for a licensed AI music platform. Warner settled with Udio too. Suno settled with Warner in November 2025 and licensed BMG afterward, but it's still fighting Universal and Sony on fair use grounds in Boston, with fact discovery closing September 30, 2026 and dispositive motions not due until April 9, 2027. Sony's claims against Udio also remain live, and Sony filed a second suit in July 2026 covering more than 30,000 additional recordings.

So the scoreboard on launch day looks like this.

Suno v6 is the more aggressive position by a distance, because it ships a fully licensed commercial product while simultaneously contesting the two largest plaintiffs in the case, which means the model sitting in your browser right now was deliberately built to exclude the exact catalogues still being fought over in a Boston courtroom. That's a legal firewall expressed as a training decision.

There's a cost to it. Suno v6 was trained without Universal and Sony recordings, and those two catalogues between them account for an enormous share of recorded music history, which is a constraint no amount of research breakthrough entirely erases. Whether that shows up as a real quality ceiling is unknowable from launch-day listening, and Brody's line is that the gains came from research breakthroughs rather than from the licensed data itself. Worth watching. If Suno v6 output ends up feeling narrower in certain genres, the missing catalogues are the first place to look.

A separate wrinkle sits under all of it. In June 2026 the American Federation of Musicians sued Universal and Warner, alleging member recordings were licensed into AI deals with no compensation, credit or disclosure. The labels licensing Suno may not have had clean authority from the musicians on those recordings. That case doesn't touch Suno v6 directly. It could reshape what a licensed AI music model even means.

Six things, before you generate anything serious.

Check what happened to your existing Custom Models. If you built one on v5.5, confirm whether it survived the transition and whether it now runs on the v6 base. Suno hasn't documented this. Your own account is the only reliable source.

Turn Variety off before you touch your prompt library. It rewrites your style prompts automatically, and the rewrites reportedly come back shorter and weaker. If you've spent months tuning prompts, this setting will undo that work silently.

Test v6-mini against Suno v6 on the same prompt. Free model, no credit risk on the comparison itself, and the launch-day reports suggest the gap is smaller than the pricing implies. If you can't hear a difference on your genre, that's a real finding about your subscription.

Count your actual releases from the last three months. Not generations. Releases. If the number is under 27 a month you belong on Pro, and if it's over you're burning about a hundred dollars a month by staying there.

Open Studio if you're on Premier. The unmetered stem export is the feature you're paying for and probably not using.

Budget in downloads. Credits are a drafting allowance and they'll never be your constraint. The 20 or 60 downloads are the product.

Free, on v6-mini, if you're testing. Seven lifetime downloads is enough to learn whether Suno v6 fits how you work, and v6-mini reportedly holds up better against the paid models than the marketing gap suggests.

Pro at $10, or $8 annually, if you release fewer than 27 tracks a month. That covers almost every independent artist. You get Suno v6, v6-wild, Custom Models, commercial rights, and 20 downloads. Draft as much as you like. You'll only ever ship 4% of it.

Premier at $30, or $24 annually, if you cross 27 releases a month, or if you're finishing tracks in a DAW. The unmetered 32-bit/48kHz Studio export is the strongest argument for Premier and it barely gets mentioned. For a producer pulling stems out to mix elsewhere, it's the only export route that doesn't count against the cap.

The trap to avoid is upgrading for credits. Premier gives you four times the credits of Pro and three times the downloads, and since downloads are the real limit, you're paying triple for a 3x increase in the thing that matters while getting a 4x increase in the thing that doesn't. Nobody upgrades to Suno v6 Premier because they ran out of credits, or at least nobody should.

Annual billing is the one obvious win on either tier. Pro drops from $10 to $8 a month and Premier from $30 to $24, a 20% cut for a commitment you were probably making anyway. On Premier that's $72 a year, which is 24 extra downloads at the $2.99 rate.

When did Suno v6 launch? September 9, 2026. Suno released v6, v6-wild and v6-mini the same day and began retiring all previous models immediately.

Is Suno v6 free? Partly. v6-mini is free for all users. The flagship Suno v6 and v6-wild require a Pro subscription at $10 a month or Premier at $30 a month.

What was Suno v6 trained on? Licensed catalogue from Warner Music Group, BMG, and participating Believe and TuneCore artists, plus user data. It was trained from scratch and excludes Universal Music Group and Sony Music Entertainment recordings.

Can I still use Suno v5.5? No. Every model before Suno v6 is being retired, with no rollback option.

How much does a Custom Model cost in Suno v6? 100 credits, which is 20 standard generations. You need at least six of your own tracks, and you can hold three models at once on Pro or Premier.

How many songs can I download on Suno v6? 20 a month on Pro, 60 a month on Premier, and 7 total for the lifetime of a free account. Extra downloads are $2.99 each with no bulk discount.

Can I sell music made with Suno v6? Yes on Pro and Premier, which carry commercial rights, and you keep those rights after cancelling. You have to spend a download first. Free-tier downloads carry no commercial rights.

Can Suno v6 tracks go on Spotify? Yes. Tracks from the industry-partner model are eligible for distribution through Believe and TuneCore, which covers Spotify, Apple Music, Amazon Music and YouTube.

Does Suno v6 watermark songs? Yes. Every output carries audio watermarking and fingerprinting via Audible Magic and Musixmatch, designed to be tamper-resistant and inaudible.

Is v6-wild better than Suno v6? No, and it isn't meant to be. It's deliberately less tuned to average preference, so it produces more unusual output and fails more often. Early hands-on testing reported roughly half of v6-wild generations coming back unusable.

Is Suno v6 actually better than v5.5? On features, clearly yes. On raw audio quality the launch-day evidence is thin, and at least one hands-on tester reported not being able to reliably tell the two apart.