Cloud Container Instance (CCI)

Deploy containerized workloads — DeepSeek, Jupyter, custom images — in seconds

Cloud Container Instance (CCI) runs any container by the second. One command stands up a DeepSeek inference endpoint; another spins up a Jupyter dev environment. Scale up, scale to zero, only pay for runtime.

Typical scenarios

LLM inference services (DeepSeek, Qwen, LLaMA…)
Jupyter / VS Code Server remote development
Batch inference and offline scoring
Ad-hoc training runs from a custom image

Get started

DeepSeek deploy

One-click DeepSeek-V3 inference endpoint

Jupyter env

GPU-enabled Jupyter Notebook image

SSH access

SSH into a container for interactive debugging

Advanced recipes

End-to-end CCI walkthroughs live in the Cookbook.

Fine-tune llama3-8b with CCI + LLaMA Factory

Single-card H800A LoRA SFT, with Webui

Deploy Baichuan2-7B-Chat with CCI + vLLM

OpenAI-compatible inference, verified with curl

ML end-to-end on CCI

Data prep → training → inference, full path

CCI vs ALab vs VKS

ALab — graphical workspace, fastest start for newcomers.
CCI — container granularity, ideal for serving, batch, and custom images.
VKS — full Kubernetes cluster for multi-service orchestration.

Billing

Per-second container runtime billing; image registry and attached storage billed by capacity.