HYOR AI Tech Ltd. · Seoul
The largest open AI models, on fleets of small machines.
HYOR AI Tech is a Seoul company. We make SparkPipe, a source-available inference engine written in C and CUDA that splits models of a terabyte and more across a fleet of NVIDIA DGX Sparks. We rent whole fleets by the hour, so a business's data stays in Korea.
SparkPipe
What we make.
Fleet-hours
Your own AI fleet in Korea, by the hour.
Sixteen DGX Sparks in Seoul serve only you for the hours you buy: Kimi K3, GLM 5.3 and DeepSeek V4.1, and Korea's K-EXAONE 2.0, A.X K2 and Solar Open 2. Sales open 1 December 2026.
sparkpipe.ai →Source available
The whole engine is public.
The kernels, the model drivers, the transport between nodes and the serving stack are on GitHub under the Functional Source License with an added permission: any use, commercial included, except renting out fleets or selling inference on them. Each version becomes Apache License 2.0 two years after release.
github.com/sparkpipe →Company