Mureka Just Got Serious. The Testing Hasn't Caught Up.
Mureka shipped V9.5 on August 13th. Five days before I'm writing this. Mureka is calling it the new flagship, and every piece of marketing copy around it says the same thing: more natural, more expressive, better than what came before.
Here's what I found when I went looking for independent testing on it: I couldn't find any. V9.5 doesn't show up yet on the Artificial Analysis Music Arena, a blind-vote leaderboard that has tested V8 and V9, and on those, Mureka sits second, behind Suno V5.5, not first. As far as I can tell, the model everyone's currently excited about, including me, doesn't have neutral evidence behind it yet. The models it just replaced do, and they lost.
I'm telling you that before I tell you anything else, because it's the kind of thing most "Mureka just got a huge update" posts won't say.
What I can actually verify right now
The generation count changed, and it breaks the math in my own book. The current Unlock Mureka defines one generation as one attempt returning one track. That's no longer true. Mureka's own Easy Mode documentation now states each generation produces two song variants, and the API defaults to two with a maximum of three. If you're planning ten generations against a Gold budget using the numbers in my book, you're actually planning twenty candidates. Every cost estimate and every case study in the current edition is wrong because of this one change.
Pricing is confirmed, with a real catch. Straight from Mureka's own site as of today: Pro is $10 a month for 5,000 Gold, up to 500 songs or 250 minutes of speech. Premier is $30 for 20,000 Gold, up to 2,000 songs or 1,000 minutes. Premier is also where DAW control, 12-stem splitting, Studio, and professional voice cloning live. The catch: Apple's in-app purchase pricing doesn't match the web pricing, different numbers entirely depending on where you subscribe. If you're comparing plans, check the web page directly, not the app store listing.
The Ableton integration isn't a plugin, and I can tell you exactly how it works even though I don't run Ableton myself. It's a control-surface remote script plus a Max for Live device, set up in three steps inside Live's settings. Worth knowing if you're on Ableton Standard rather than Suite: Max for Live is a paid add-on outside of Suite, so DAW control may not be available to you at all depending on which Ableton license you have. Mureka doesn't say that anywhere in their own materials. It's a consequence of the requirement, not a documented limitation.
Stem splitting now goes up to 12 stems plus MIDI, not the 4-stem ceiling the current book describes. Real numbers from someone who's actually paid for it: 100 Gold for a 12-stem-plus-MIDI pull, 20 Gold for a basic two-stem split.
What's still just a lead worth testing, not a finding
The book's current rule on structure tags is absolute: the bare `[Guitar Solo]` tag fails, always use `[Instrumental Break]` instead. A community report from earlier this year suggests a qualified version, something like `[8 measure guitar solo]`, gets recognized some of the time under the newer models, while still failing other times. That's one thread, not a lab result. Worth trying yourself. I'm not rewriting the book's rule on it yet.
Why the book's getting rebuilt, not patched
Every one of the specifics above breaks something specific in the current edition: the cost math, the stem count, the pricing tiers, the model lineup. That's not a chapter that needs an update note. It's a full rewrite.
I'm not rushing it out before it's tested the way everything else I publish is tested, which means real generations, real comparisons, not marketing copy repeated back with my name on it.
If you buy Unlock Mureka now, you'll get the updated edition free when it ships. One-time offer tied to this version transition, not an ongoing promise. Once the rebuilt edition is live, this closes with it.
If you're deep in Mureka right now and you've found something that contradicts what I've said here, tell me. The last thing I want is to publish a rebuilt book off the same kind of unverified confidence I just spent this whole post pushing back on.
Josh
Founder, JG BeatsLab
