Nunchux launches: one API aggregating 30+ image, video and avatar models

junyanz89 · x · 2026-09-04

Founder lmxyy1999 traces seven years of efficient generative AI research — from GANs to diffusion to world models — arguing models must be fast, cheap and reliable to be practical. That path led to the Nunchaku engine and now Nunchux.

Nunchux goes live targeting multimodal generative AI inference: fast, affordable, high-quality serving for image, video and world models. Its Modelverse puts 30+ image, video and avatar models behind one API, with Radical Speed and Radical Value tiers plus partner APIs. Waitlist is open with $10 credits for new accounts; the team is hiring.

Original post →

More from Infra

Infra channel →