Sand.ai

Sand.ai is an artificial intelligence research and product company headquartered in Beijing, China, specializing in generative video models. The company was founded in January 2024 by Cao Yue, a prominent researcher known for co-authoring the Swin Transformer, a foundational paper in computer vision. Yue's extensive background includes leading a major AI research institution's Vision Model Center, where he focused on fundamental vision and multimodal large models, accumulating over 50,000 academic citations. The firm has distinguished itself by pursuing an autoregressive approach to video generation, a less common path than the mainstream Diffusion technique. Its flagship model, Magi-1, is open-source and has demonstrated high performance on physical realism benchmarks, outperforming many peers. Sand.ai follows a dual strategy of developing core models and commercial products. Its music agent product, VidMuse, launched in January 2026 and achieved an annual recurring revenue (ARR) of $10 million within two months. Sand.ai operates with a lean team of under 30 members and has a strong focus on open-source collaboration to accelerate progress in the field. The company recently secured over $100 million in funding across two rounds from a syndicate of notable investors, including Look Capital, Lollapalooza Capital, IDG Capital, and Baidu Ventures. The capital is intended to support the development and launch of its next-generation video model, which will be based on a Mixture-of-Experts (MoE) architecture and is also planned for open-source release. Keywords: video generation, generative AI, autoregressive model, Mixture-of-Experts, MoE, open-source AI, computer vision, multimodal models, machine learning, deep learning, AI video, Magi-1, VidMuse, Swin Transformer, AI research, visual models, AI models, Cao Yue, artificial intelligence video, physics-based AI