AI inference product matrix for individuals and enterprises
NVIDIA Jetson Thor Module
Runs 230B MoE Quantized Models Locally
2070 TFLOPS (FP4)
Supports Multi-Bay Storage Expansion
AMD Ryzen AI Max Series
Innovative Honeycomb Liquid Cooling
Moore Threads SoC Chip
Palm-sized Ultra Compact Design
MetaX Tech N260 Chip
Full Domestic Stack & Security Compliant
Single-cabinet integration, lower deployment costs
40 cards, 5 nodes
Three-cabinet modular deploy
80 cards, 10 nodes
70kW cooling capacity
128 cards, 16 nodes
Cluster-scale high density
13000 TFLOPS (FP8)
8 x GPU high-density compute
