MI
Foundation Models
MiniMax
Open-weight multimodal models spanning text, voice and video
MiniMax is a Shanghai-based, Hong Kong-listed startup building foundation models across text, audio and video. Its M-series models are open-weight; M3 (2026) combines long context, native multimodal input and computer use.
Models
- MiniMax M3 — Open-weight flagship: 1M context, multimodal, computer use (2026)💬 Text🧮 Reasoning💻 Code👁️ Multimodal
- Hailuo (video) — Text-to-video generation🎬 Video
- MiniMax Speech / Music — Voice cloning and audio generation🎙️ Audio
Highlights
- M3 (2026): first open-weight model to combine long context, multimodal input and computer use
- Known for high-quality voice and video generation products
Information compiled from public sources; refer to the company's official website for authoritative details.