What it does
MiniMax develops multimodal foundation models covering text, audio, image, video and music, and offers them to developers and enterprises through an open platform. Its language line includes MiniMax M3, M2.7 and M2.5. M3 is presented as a coding and agentic model with a one-million-token context window, a sparse attention design the company calls MSA, and native multimodality. Other model families are MiniMax H3 for video generation, MiniMax Speech 2.8 for speech and MiniMax Music 3.0 for music. The site presents H3 and Music 3.0 as open models, and Music 3.0 is labelled open-weights. MiniMax also runs its own products: MiniMax Code, a coding harness built for its models with a Coding Mode and a Work Mode; MiniMax Design, an agent platform for commercial content such as ads, e-commerce assets and brand campaigns; MiniMax Audio; and Talkie. Platform documentation for text generation, video generation and voice cloning is linked from the home page. The company says it was founded in early 2022.
Capabilities
- MiniMax M3 language model with a 1M context window and a coding and agentic focus
- MiniMax H3 video generation model
- MiniMax Speech 2.8 and Music 3.0 for voice and music
- API access through the MiniMax platform
- MiniMax Design agent platform for ads, e-commerce assets and brand campaigns
- MiniMax Code, a coding harness built for its models
When to use it
- You want a hosted model API for coding, agent or long-context work
- You need video, speech or music generation from a single provider
- You are interested in open-weight video or music models
- You create ad or e-commerce content with an agent platform
