MiniMax

MiniMax's H1 2026 interim report records US$73.9 million in Open Platform and enterprise AI revenue, versus US$9.2 million in H1 2025, representing 63.4% of total revenue. This provides monetized API and enterprise adoption evidence distinct from user-count claims.

MiniMax reports more than 300 million users across 230+ countries and regions and more than two million enterprises and developers, powered by multimodal foundation models and multiple consumer, developer and enterprise products.

MiniMax reports more than 300 million users across 230+ countries and regions and more than two million enterprises and developers, powered by multimodal foundation models and multiple consumer, developer and enterprise products.

MiniMax released H3 on July 31, 2026 and documents Contextual Omni Representation, H3-VAE, H3-Omni Transformer and in-context regeneration for multimodal video/audio generation. These are distinct technical contributions from the existing M3 coding-model evidence.

MiniMax M3 combines frontier coding, agentic performance, 1M-token context and native multimodality using MiniMax Sparse Attention, representing strong original model architecture and broad capability.

MiniMax M3 combines frontier coding, agentic performance, 1M-token context and native multimodality using MiniMax Sparse Attention, representing strong original model architecture and broad capability.

MiniMax's annual-results announcement documents its January 9, 2026 Hong Kong listing and net proceeds of approximately HK$4,596.1 million. The completed capital-market event is independent of the existing 2025 revenue and international-customer evidence.

MiniMax operates a global consumer and enterprise platform, generated $79M revenue in 2025 with more than 70% international, and serves users, enterprises and developers across more than 100 countries.

MiniMax operates a global consumer and enterprise platform, generated $79M revenue in 2025 with more than 70% international, and serves users, enterprises and developers across more than 100 countries.

MiniMax provides M1 model weights and technical reports via its official Hugging Face and GitHub accounts, with inference support in vLLM and Transformers and collaboration on SGLang support. These reusable artifacts and serving integrations support inferred skill portability.

MiniMax works across multimodal foundation models, coding, agents, long-context systems, speech, video, music, APIs and global product deployment, producing broadly transferable frontier-AI skills.

MiniMax works across multimodal foundation models, coding, agents, long-context systems, speech, video, music, APIs and global product deployment, producing broadly transferable frontier-AI skills.