Nunchux AI is building the world’s fastest and most cost-effective inference infrastructure for image, video, and world models—full-stack software that pushes generative media beyond the limits of today’s hardware.

Its Modelverse platform brings 30+ image, video, and avatar models behind a single API, letting developers and enterprises run generative media up to 100x faster and 10x cheaper with no loss in fidelity.

Nunchux was founded by the MIT researchers behind SVDQuant, QServe, and Sparse VideoGen, alongside veterans from NVIDIA, AMD, and Snowflake. The company is headquartered in San Francisco, California.