LLM runtimes
Serve open-weight models behind your own private API.
An LLM runtime serves open-weight models behind an API you control. On CPU plans, quantized 7-8B models run at roughly 5-15 tokens per second — usable for chat, retrieval and automation, not for realtime products. Ollama is the usual starting point; llama.cpp and KoboldCpp trade convenience for control.
A runtime is the piece that actually loads model weights and answers requests. Everything else in the catalogue — chat interfaces, RAG tools, automation steps — talks to one of these or to a commercial provider. Running your own means no metering, no rate limit that is not yours, and no prompt leaving the machine.
The honest constraint is speed. These are CPU plans, and CPU inference is fast enough to be genuinely useful and slow enough that you should know before you buy. The full picture, including which model sizes fit which plan, is on the private LLM hosting page.
LLM runtimes we host
Ollama
Your private LLM server — pull a model, get an API, keep every token on your own box.
from $16.99/moDetails →
Hermes
Nous Research’s Hermes models, served privately behind your own API.
from $16.99/moDetails →
LocalAI
A drop-in OpenAI replacement for text, images, audio and embeddings.
from $16.99/moDetails →
KoboldCpp
A single-binary GGUF server with its own UI and a KoboldAI API.
from $13.99/moDetails →
llama.cpp
The reference GGUF inference server, with nothing between it and the CPU.
from $13.99/moDetails →
Text Generation WebUI
Oobabooga’s all-in-one interface for running and testing local models.
from $16.99/moDetails →
vLLM
High-throughput model serving — GPU territory, listed for completeness.
on requestDetails →Choosing between these
Straight answers
Ollama or llama.cpp?
Can these serve several applications at once?
Put your AI stack on your own box
Pick an app, pick a size, and have it running today. Month to month, cancel whenever.