Self-Hosted AI Agent
A small black Lenovo ThinkCentre Tiny standing next to a coffee mug

Can Codex run on a Raspberry Pi 5?

Not confirmed. Codex’s requirements list Ubuntu and Debian but don’t say whether ARM64 is supported. Check its release downloads before buying a Pi for it.

What Codex needs

From Codex’s own documentation, checked 28 October 2026.

Operating systemmacOS 12+, Ubuntu 20.04+ or Debian 10+, or Windows 11 via WSL2
Memory4 GB minimum, 8 GB recommended
ProcessorThe requirements table does not name a processor architecture

What a Raspberry Pi 5 brings

Power and running costAbout 2.7 W at idle, so roughly $0.36 a month left on all day at the US average electricity price
Good to knowOpenClaw's docs rate a 4 or 8 GB Pi 5 as the best Pi for its gateway, recommend a USB SSD over an SD card, and advise against running local models on a Pi.
Left on all dayThe lowest power draw here, but it is ARM-based, so some optional tools may lack builds, and it is too slow for local models.

Local models on a Raspberry Pi 5

OpenClaw’s Raspberry Pi guide says not to run local models on a Pi: “even small models are too slow to be useful.” Use a cloud model. The memory table shows what each model size needs.

Questions

Is a Raspberry Pi 5 enough for Codex?

Not confirmed. Codex’s requirements list Ubuntu and Debian but don’t say whether ARM64 is supported. Check its release downloads before buying a Pi for it.

What does it cost to keep a Raspberry Pi 5 running for Codex?

About 2.7 W at idle, so roughly $0.36 a month left on all day at the US average electricity price. The AI model is usually the bigger monthly cost.

Can Codex use a local model on a Raspberry Pi 5?

OpenClaw’s Raspberry Pi guide says not to run local models on a Pi: “even small models are too slow to be useful.” Use a cloud model.

Other combinations

Codex on An old laptop · An old office PC · An N100 mini PC · A Mac mini · A VPS.

On a Raspberry Pi 5: OpenClaw · Hermes Agent · Claude Code. See every agent and computer.

Sources