Foundation models for text, audio, video, and agents — all from one Chinese AI lab. Train, deploy, and scale on the MiniMax platform.
Reasoning model with 128k context. 12% faster inference, 38% lower price-per-token than M2.6.
Full-song generation in 32 styles. Instrumental separation and lyric-aware composition.
10s 1080p text-to-video at 24fps. Cinematic motion control and camera-aware prompts.
28 languages, 480 voices, sub-200ms streaming. Emotion control via prompt tags.
Build agents, generate media, run open weights — through one SDK and one billing plan.
| Model | Context | Modality | Status | Price / 1M tok |
|---|---|---|---|---|
| M2.7 | 128k | Text · Reasoning | NEW | $1.20 / $4.80 |
| Music 2.6 | — | Audio · Music | Available | $0.04 / second |
| Hailuo | — | Video · 1080p | BETA | $0.12 / second |
| Speech 2.8 | — | Audio · TTS | Available | $0.018 / 1k chars |
| M2-Lite | 32k | Text · Open Source | Available | Self-host |
Run M2.7 in production with confidence. Active until June 30 — credits land monthly.
Tokens, type, components — the chrome that carries every MiniMax surface.
| Token | Sample | Spec |
|---|---|---|
| hero-display | Intelligence with everyone. | DM Sans · 80 / 600 · -2px |
| display-lg | Models built to ship. | DM Sans · 56 / 600 · -1.5px |
| heading-lg | Pick the right weight class. | DM Sans · 40 / 600 · -1px |
| card-title | Hailuo Video | DM Sans · 20 / 600 |
| body-md | Foundation models for text, audio, video, and agents. | DM Sans · 16 / 400 · 1.5 |
| caption-bold | BETA / NEW / AVAILABLE | DM Sans · 13 / 600 |
| button-md | Start building | DM Sans · 14 / 600 |