Self-hosted inference orchestrators compared: LocalAI, exo, GPUStack, vLLM
What happened
A reference comparison of the self-hosted AI orchestrators in 2026: modalities, multi-machine support, auto-discovery, cache-aware routing, ops console, cloud burst, non-LLM fan-out, training, Kubernetes, platforms and signed images — with a pick-by-situation guide and sources.
Summary assembled by rule from the sources below