HiDream.ai raises RMB 1.5 billion and says its unified multimodal stack can rival 56B-model systems

创业邦 · wechat · 2026-07-24

HiDream.ai raises RMB 1.5B as it pushes a unified multimodal path toward world models

Chinese AI company HiDream.ai (智象未来) closed a RMB 1.5 billion Series C led by multiple state-backed and industrial investors, bringing its three-month fundraising total to more than RMB 2.1 billion and making it a unicorn.
The article argues that HiDream is betting on a native unified multimodal architecture (UiT) instead of the usual text-plus-image/video stack. It claims the architecture can map pixels, text tokens, video voxels, audio, motion and spatial relations into one shared token space, improving controllability and reducing information loss.
The company says its 8B model can match the output of a traditional 56B model, and highlights product milestones such as HiDream-O1-Image-Pro (200B+ parameters), HiDream-I1 topping a major ranking within 24 hours, and vivagoR1, a long-form multimodal content agent that supports continuous video generation.
Commercially, HiDream says it passed RMB 100 million in annual revenue in 2025, signed more than RMB 400 million in Q1 2026, serves 50 million+ users and 40,000 enterprise customers, and operates in 100+ countries.

Competitive context

The piece also compares HiDream with rivals such as Vidu, PixVerse, and Google’s multimodal efforts, framing the race as a shift from pure generation to perception, simulation, and action.

Original post →

HiDream.ai raises RMB 1.5 billion and says its unified multimodal stack can rival 56B-model systems

More from Venture