📌 Pinnedreleaseinferencecontrol-planescheduling
Modelplane v0.6: making the most of your hardware
Modelplane v0.6 packs small models onto shared GPU nodes and serves on AMD GPUs and CPUs as well as NVIDIA. SGLang prefill/decode works too, behind a router you write.