System Requirements#

Hardware#

Component

Requirement

Processor

Intel® Core™ Ultra, Intel® Core™, or Intel® Xeon® processor (x86-64)

GPU

Intel® GPU (integrated or Intel® Arc™ discrete) exposed at /dev/dri/renderD*

Memory

At least 64 GB for mixture-of-experts (MoE) models such as Qwen3-30B-A3B (the default); check the OpenVINO™ defaults and the model card for the model you choose. At least 32 GB for Qwen3-8B or a similar dense model. Integrated GPUs share this system memory for the model weights and the 4 GB KV cache

Disk

About 30 GB free: the INT4 model files (about 5 GB for Qwen3-8B, about 16 GB for Qwen3-30B-A3B), plus container images and data

Software#

Component

Requirement

Operating system

Linux x86-64 with Intel GPU drivers installed

Docker

Docker Engine with the Compose v2 plugin, usable by the current user without sudo

Tools

Git, make, curl, Python 3 (used by the model download script)

Network

Access to GitHub, Docker Hub, and Hugging Face during installation

Verify the System#

Confirm the Intel GPU device is visible:

ls /dev/dri/renderD*

Confirm Docker works for your user:

docker run --rm hello-world
docker compose version

If Docker requires sudo, add your user to the docker group, then log out and back in:

sudo usermod -aG docker "$USER"

Network Ports#

Port

Service

Default binding

8600

Operator UI and event webhook

All interfaces (0.0.0.0)

4444

OpenVINO™ Model Server API

Loopback only

8090

Weather service

Loopback only

See how to change ports and bindings in Configuration.


Back to Get Started