AN ENGINEERING REFERENCE FOR MODEL SERVING
From model
to machine.
Find a deployment recipe that fits your hardware. Configure the engine, inspect the command, and run it on your terms.
CURATED FOR REAL HARDWARE / UPDATED FROM SOURCE RECIPES
LIVE INDEXMODEL → HARDWARE → COMMAND
01 / MODEL
02 / HARDWARE
03 / ENGINE
04 / COMMAND
TARGET HARDWAREH200196 compatible recipes
EXAMPLE RECIPE / vLLMflux-3-image-klein
Browse 196 01 / COMPUTE INDEX
Start with your hardware.
Choose the accelerator you already run. The index narrows to recipes with a compatible checkpoint and serving path.
02 / RECENT ADDITIONS
Fresh from the index.
03 / IMAGE REFERENCE
One stack.
The right images.
Resolve GPUStack server and worker images for your accelerator, runtime, architecture and registry.