Music quality is difficult to describe with a single number. Listening quality, vocals, sound quality, prompt control, genre expression, and audio health all measure different problems.
Mureka V9.5 was evaluated across these dimensions using the same 100-song test set: 35% overall listening pass rate, 61% vocal pass rate, 35% average sound-quality pass rate, 97% prompt-control pass rate, 95.7% full genre expression, 84% audio health, and 28% combined usability.
Just as importantly, the evaluation preserves the trade-off. V9.5 was not highest in the average music-quality category because it did not choose the fullest arrangement strategy. That boundary matters. A useful benchmark should not only tell you where a model wins, but also what it chooses to trade off and why.
Ready to create more natural, controllable AI music? Try Mureka V9.5.
Top comments (0)