Mureka released V9.5 of its AI music model on August 31, 2026, shifting the focus from generating convincing sounds toward building complete songs with stronger arrangements, more natural vocals, and clearer musical direction. The company also previewed an upcoming O3 model that adds test-time scaling on top of the same foundation.
Try It: Generate a Full Song With Structured Sections
V9.5 is built to improve how sections, instruments, and vocal parts develop across an entire track rather than simply stacking more layers. In practice that means you can prompt for a song with a real intro, verse, and chorus and get more coherent transitions and dynamics between them. If you produce, the natural next step is to export a generated stem or section into a DAW like Ableton and build around it, using Mureka for the vocal or arrangement idea and your own tools for the mix.
Why It Matters for Creators
The gap between AI music that sounds impressive for eight seconds and AI music you can actually finish a track with has been the whole problem. By targeting phrasing, arrangement, and the interaction between lead vocals and the surrounding production, V9.5 aims squarely at that gap. For content creators who need original, cleared music under video, a model that produces coherent full songs is more useful than one that produces loops.
Key Details
Vendor-reported metrics. In Mureka's internal testing, V9.5 produced lead vocals that listeners considered convincing in 61.0 percent of cases, followed user prompts in 97.0 percent of generations, and matched requested genres 95.7 percent of the time. These are self-reported numbers, not independent benchmarks.
Arrangement focus. The update emphasizes dynamics, space, and transitions across a track, aiming for cohesive songs rather than layered stems, as Fame Magazine noted.
O3 preview. Mureka also previewed O3, which is not yet released. It applies test-time scaling to give the model more room to reassess a song as it develops.
What to Do Next
Open the Mureka platform, prompt a full song in a genre you actually use, and judge the arrangement and vocal coherence against your last AI music tool. Coverage from the launch announcement outlines what changed if you want the full breakdown before testing.
