Is MiniMax Music 3.0 available in the API?

Yes, but on the China endpoint. As of July 2026, music-3.0 is callable on MiniMax's China API (api.minimaxi.com, the 海螺 platform) with a matching China key. The international API (api.minimax.io) still tops out at music-2.6. Many wrapper sites either falsely imply international 3.0 access or quietly run 2.6.

What does MiniMax Music generate?

Full songs with vocals and layered instrumentation from a style prompt and lyrics, pure instrumentals, or AI covers that restyle an uploaded track while keeping its melody. Output is up to about 5 minutes, in 40+ languages, commercially licensed.

Does MiniMax Music have an official API, unlike Suno?

Yes. MiniMax offers an official public music generation API (api.minimax.io internationally, api.minimaxi.com in China). Suno has no official API, so every 'Suno API' is an unofficial relay. For developers who need a supported first-party endpoint, this is MiniMax's clearest advantage.

Model guide

MiniMax Music 3.0, explained straight.

MiniMax Music turns a style prompt and lyrics into a finished song, vocals and all. Version 3.0 is real, and it is callable from the API right now — but there is a region catch that almost every site ranking for the term gets wrong. Yif covers the audio models worth knowing; here is the honest version, with a track we generated on 3.0 to prove it.

The region catch, up front

“Music 3.0” is the new flagship, and yes, it is in the API — but only on MiniMax’s China endpoint. Your key and the host region have to match.

China API (海螺)api.minimaxi.com — music-3.0 works here (verified: we generated the track above). Needs a China-region key

International APIapi.minimax.io — tops out at music-2.6. No music-3.0 string yet. Needs an international key

Consumer appminimax.io/audio — shows Music-3.0 as the default “New” model, credits shared across music and voice

So a tool advertising “Music 3.0” is telling the truth only if it runs on the China stack. On the international API it is really 2.6. That distinction is the thing to check, and the thing the squatter pages skip.

What it generates

Full songs with vocals and layered instrumentation from a prompt plus lyrics, up to about 5 minutes.
Pure instrumentals (set is_instrumental), no lyrics needed — that is what the sample above is.
AI Cover: upload a track, keep its melody, restyle it (the music-cover models).
Lyrics Optimizer: leave lyrics blank and it writes them from your style prompt.
Structure tags in the lyrics: [Intro] [Verse] [Chorus] [Bridge] [Outro] and more. 40+ languages.

How to access it

China APIapi.minimaxi.com, music-3.0. Self-serve signup at platform.minimaxi.com, no enterprise gate; new accounts get trial credits

International APIapi.minimax.io, music-2.6 at $0.15 per song (up to 5 min, currently limited-free); music-2.0 at $0.03; a free-tier music-2.6-free works on any key at lower rate limits

Consumer appminimax.io/audio — the easiest way to try 3.0 with no code, on shared subscription credits

The request shape

Synchronous POST that returns the finished song as hex-encoded audio (or a 24-hour URL). No task polling — the call holds open until the song is done, which takes a couple of minutes.

POST https://api.minimaxi.com/v1/music_generation
Authorization: Bearer <CHINA_API_KEY>
Content-Type: application/json

{
  "model": "music-3.0",
  "prompt": "Dreamy synthwave, nostalgic, night drive, warm analog pads",
  "is_instrumental": true,
  "output_format": "url",
  "audio_setting": { "sample_rate": 44100, "bitrate": 256000, "format": "mp3" }
}

// response.data.audio is the track (status 2 = done);
// response.extra_info.music_duration is its length in ms

prompt: up to 2,000 chars of style, mood and scene.
lyrics: up to 3,500 chars with structure tags; required for vocal tracks unless lyrics_optimizer is on.
Billed per song (up to 5 min), not per second — a short clip costs the same as a full track.

MiniMax Music vs Suno, briefly

Both make full songs, so they get compared. Suno’s latest generation is widely rated ahead on raw fidelity and vocal realism. MiniMax’s edges are a real official API (Suno has none, so every “Suno API” is unofficial), predictable and controllable output, and a strong reference-audio cover mode. If you are building software and need a first-party endpoint, that API gap is the deciding factor.

Generate on Yif

Music generation is on the way to the Yif playground on the same honest terms, running music-3.0. One engineering note we will not hide: a full song takes two to four minutes to render, which is longer than a single web request can stay open, so it needs an async pipeline rather than a live wait. That is being wired now. In the meantime, hear what one-pass audio scene generation sounds like in the Seed-Audio playground, or read how Seed-Audio and Suno split the work in the scene-vs-song comparison.