What it does
The MiniMax API Platform is where developers reach the company's models over an API. Its models overview groups them into language, video, audio and music families, each with a current set and a legacy set. Language models run from MiniMax-M3, described as a frontier multimodal coding model with a 1M context window, through M2.7 and a faster M2.7-highspeed variant, down to legacy M2.5, M2.1 and M2 releases. The page gives M2.1 as 230B total parameters with 10B active per inference, and M2 as a 200k-token context with up to 128k tokens of output. Video covers MiniMax H3, which takes text, images, first and last frames or multimodal references and produces 4 to 15 second clips at 24 fps in 768P or 2K, and H3 Max, a faster variant post-trained by fal.ai that outputs 480P or 768P. Hailuo 2.3 and 02 are listed as legacy. The speech models speech-2.8-hd and speech-2.8-turbo support 40 languages and 7 emotions. Music generation lists music-3.0, music-2.6 and music-cover. A notice says that from August 20, 2026 the paid music and lyrics APIs close to new users, existing paying users can continue, and the free music APIs are discontinued. It points to MiniMax Audio and the open-source Music 3 model on Hugging Face instead.
Capabilities
- Language models up to MiniMax-M3 with a 1M context window
- H3 video generation from text, images, first and last frames or references
- Speech models covering 40 languages and 7 emotions
- Music models including music-3.0 and music-cover
- Legacy versions kept listed beside the current ones
When to use it
- You want to call MiniMax language, video, speech or music models from your own code
- You are choosing between current and legacy MiniMax model versions
- You need video clips of up to 15 seconds from text or reference images
