What's up with Suno and the others these days?
Taking a look at modern AI music generation models and tracking their futures
Take a guess at how many AI music models have been released this year.
Following Suno and Udio’s deals with the majors, I think many in the industry anticipated a slower cadence as the companies making these models seek to make fairly licensed models.
This is not the case.
So far in 2026, 13 different prominent AI music generation models have been released. These span across 8 different providers, including prominent companies like Suno and Google. As listeners seem to be growing more and more disinterested in the AI music that’s crowding DSPs, I found this surprising. In the same timeframe in 2025 there were only 11 models released. If anything, the frequency of AI music model releases is speeding up.
Keeping track of all the AI music models being released is a challenge, so I published a page on my site to help folks. You can scroll thorough the history of this space and listen to samples from each model to see how far they’ve come in quality and fidelity.
Here are some tidbits I’ve noticed looking at the current landscape:
Suno charges forward. Despite their deal with the majors that many (including myself) thought would indicate a pause in model development, they’ve continued on. Suno 5.5 seems to be following their standard release cadence, with no mention in the release notes on the data used to power it. The new model does place an emphasis on user remixing/voice additions, which was a focus in the major partnership announcement.
Google wants everyone to generate AI music. Lyria 3 Pro is a true competitor to Suno’s dominance, acquisition of Producer.ai (and subsequent rebrand to “Google Flow Music”), and public availability of the Lyri model in their API means they have no issues with spreading high-quality AI-generated music far and wide. I anticipate many “white label” music generation services will launch this year as a result (such as ONCE Distribution’s new generation feature that seemingly uses Lyria.)
Udio is MIA. Once Suno’s largest competitor, Udio seems to be completely absent from the modern space. Their deals with the major seemed a bit more “final” in messaging, so it’s possible they just don’t have the willpower (or the user metrics) that Suno does to continue on. It probably doesn’t help that they just admitted in court that they scraped YouTube for their model.
Open source models are getting better. ACE-Step v1.5 isn’t perfect, but it’s a significant improvement over Stable Audio Open and the like. Given the amount of research papers I’ve seen in this space in the past one or two years, I anticipate these models to continue catching up (but probably always lag behind Suno, like open source LLMs lag behind GPT and Opus.)
“Fairly trained” models are far and few. I had hoped this space would continue to expand with the Suno/Udio deals and the release of Eleven Music last August (which is reportedly trained off the back of licensing deals with Kobalt and Merlin), but there haven’t been many significant fairly trained launches since. The most significant new release, Lyria 3 Pro, was likely trained off “publicly available” YouTube data, and Suno 5.5 makes no mentioned of licensed data.
Overall: the progress isn’t stopping, fairly trained models aren’t here yet, and AI music is going to continue to grow faster than we can control it.
Let’s pray the industry can successfully adapt.
Hey! Thanks for reading.
I’m at Music Biz 2026 in Atlanta this week talking about Vobile’s best-in-class AI music detection. It’s a product I’ve been working on for a while now (with the best R&D devs of all time), and one I’m very proud of.
Send me a DM to meet up and chat!