Multimodal Frontier: China’s MiniMax Releases Sora Competitor, H3 Video Model

US technological sanctions have definitively lost their meaning. On July 31, 2026, Chinese AI startup MiniMax officially introduced `H3`—an advanced multimodal video model. The algorithm works natively with text, images, and audio, generating high-quality video sequences.

This release is a harsh slap in the face to Western monopolists (OpenAI and Google). Chinese engineers are proving they can train resource-intensive generative networks for processing heavy media content even under a hardware blockade. H3 directly competes with America’s Sora in the creative industries and digital marketing segments. The expansion of the product line of Asian startups means the domestic PRC market is completely closed to Western software. Moreover, multimodal models of H3's caliber will be aggressively exported to the Global South, depriving Silicon Valley of billions in B2B revenues.

Source: MiniMax / Reuters