Moonshot AI · 1043B parameters. Trillion-param (1.04T) MoE — 32B active, 384 experts, MLA attention; among the strongest open-weights agentic/coding models. At FP16 (~2 TB) it needs a multi-GPU cluster; ships native FP8 (~1 TB).
Spanvero shows the honest price three ways — free on your own hardware, your own rented GPU at the vendor's rate, or your own API key — with $0 markup. Loading the interactive advisor…