
Dramatically cheaper than the premium tier
A low enough cost per generation that auditioning twenty variations of one lyric is a normal workflow rather than an expensive one.
MiniMax
Full songs at a fraction of the premium tier's cost per generation.
In short
Figures verified 2026-08-10. This field moves quickly — re-check before relying on them.
MiniMax matters most to anyone generating at volume, where per-track cost stops being a rounding error. The 2.5 generation closed much of the quality gap with the premium Western models while staying dramatically cheaper, which changes what workflows are viable: auditioning twenty arrangements of the same lyric is reasonable at this price and wasteful at four times it. The lyric handling is a genuine strength. It accepts long input with explicit structure tags, so a complete sheet with marked verses and choruses is followed rather than paraphrased, which is exactly what a songwriter with finished words needs. Chinese-language output is the best in the category by a clear margin, reflecting where the model was built and trained. English vocals are competent rather than exceptional — the diction is clean and the phrasing is less expressive than Suno's, which is the trade you are making for the price.
Strengths

A low enough cost per generation that auditioning twenty variations of one lyric is a normal workflow rather than an expensive one.

Long lyric sheets with explicit structure tags are followed rather than paraphrased, which is what matters when the words are already finished.

Chinese-language vocal output is the strongest in the category, which no Western-built model currently matches.
How it compares
MiniMax Music 2.5 generates complete songs with vocals from a lyric sheet and a style description, at a cost per generation substantially below the Western premium tier. It accepts long lyric input with structure tags and is unusually strong on Chinese-language vocals.
Compare with MusicGenerate
Searched as minimax music, minimax ai music, minimax music 2.5 and hailuo music. The cost-per-track searches dominate, which is the right instinct — that is where it wins.
Substantially less than the Western premium tier, which changes what workflows are viable — auditioning twenty arrangements of one lyric is reasonable at this price and wasteful at four times it.
Yes, and it is a genuine strength. It accepts long input with explicit structure tags, so a finished sheet with marked verses and choruses is followed rather than paraphrased.
It is the strongest in the category by a clear margin, reflecting where the model was built and trained. No Western-built model currently matches it there.
Competent rather than exceptional. Diction is clean and phrasing is less expressive than Suno's — that is the trade you are making for the price.
Terms are set by MiniMax's API and platform agreements, and the licensing language differs from Western platforms in ways that matter to distributors. Read the current terms before building a release on it.
Yes, and that is how most volume users reach it. Per-call pricing is what makes generating many variations of one lyric a normal workflow rather than an expensive one.
Keep exploring
More models
Models and platforms we track, compared on capability and licence.
Full songs at a fraction of the premium tier's cost per generation.