ClawBox 仕様
ボックスの中身は？
67 TOPSのAI性能、8GBメモリ、512GBストレージを備えたNVIDIA Jetson Orin Nano Super。
📅 Updated March 2026
With OpenClaw v2026.3.14 and NemoClaw from NVIDIA, hardware requirements have evolved. The minimum specs remain 4GB RAM and Node.js 20+, but for the best experience with browser automation, multi-agent workflows, and voice pipelines, we now recommend dedicated hardware with NVIDIA CUDA support. The Jetson Orin Nano remains the sweet spot — 67 TOPS of AI compute at just 15W, enough for simultaneous LLM inference, browser automation, and voice processing.
Need the recommendation, not just the specs?
This page explains minimum and recommended specs. If you want the final buying answer, read our guide to the best OpenClaw hardware, where we compare ClawBox, Raspberry Pi, Mac Mini, and other always-on AI setups.
完全な仕様
ClawBoxハードウェアについて知っておくべきすべて
AI性能
67 TOPS
NVIDIA Jetson Orin Nano Super with dedicated AI accelerator
GPU
1024-core Ampere GPU
625 MHz base clock, 40.96 GB/s memory bandwidth
メモリ
8GB LPDDR5
102.4 GB/s bandwidth, unified CPU/GPU memory
ストレージ
512GB NVMe SSD
PCIe Gen3 x4, 2,100 MB/s read speed, pre-installed
消費電力
7-25W
15W typical operation, 20W USB-C PD adapter included
接続性
WiFi 6 + Bluetooth 5.0
Dual-band 802.11ax, Gigabit Ethernet port
ポート
USB-A, USB-C, HDMI, Ethernet
microSD slot, 40-pin GPIO header
サイズ
100 × 79 × 31mm
Compact carbon fiber case, 260g total weight
AIモデル性能
ClawBoxで実行できるもの
| モデル | 速度 | 品質 |
|---|---|---|
| Llama 3.1 8B (Q4) | ~3.5 tok/s | Excellent |
| Mistral 7B (Q4) | ~4.2 tok/s | Excellent |
| Phi-3 Mini (Q4) | ~6.8 tok/s | Good |
| Gemma 2B (Q4) | ~9.1 tok/s | Fast |
| TinyLlama 1B (Q4) | ~15.3 tok/s | Basic |
| Qwen2.5 7B (Q4) | ~4.0 tok/s | Excellent |
大規模なモデルには、クラウドAPI（Claude、GPT、Gemini）でのBYOKをご利用ください。
同梱内容
ハードウェア
- ✓ NVIDIA Jetson Orin Nano Super
- ✓ 512GB NVMe SSD（プリインストール済み）
- ✓ カーボンファイバーケース
- ✓ 20W USB-C 電源アダプター
- ✓ イーサネット＆WiFi 対応
ソフトウェア
- ✓ Ubuntu 22.04 LTS
- ✓ OpenClaw プリインストール済み
- ✓ ブラウザ自動化対応済み
- ✓ 音声パイプライン設定済み
- ✓ 無料ソフトウェアアップデート
電力要件
アイドル時7W
7W
通常15W
15W
最大25W
25W
付属の20W USB-C 電源アダプター一つで十分です。
Why Dedicated Hardware for OpenClaw?
Running OpenClaw on dedicated hardware vs shared VPS or your daily-use computer
✓ Dedicated Hardware
- ✓24/7 uptime — your AI assistant never sleeps. Heartbeat checks, scheduled tasks, and proactive monitoring run around the clock without interruption.
- ✓No resource competition — CPU, GPU, and RAM are 100% dedicated to OpenClaw. No other apps fighting for memory or causing unexpected slowdowns.
- ✓Consistent performance — predictable response times for AI inference, voice processing, and browser automation. No “noisy neighbor” problems.
- ✓Lower long-term cost — ClawBox at €549 one-time vs VPS at €20-40/month (€720-1,440 over 3 years). The hardware pays for itself in under 2 years.
- ✓True privacy — your data stays on your physical hardware, on your network. No cloud provider has access to your conversations, files, or automations.
✗ Shared VPS / Your Laptop
- ✗Downtime when you close your laptop — OpenClaw stops when your computer sleeps, restarts, or runs out of battery.
- ✗Resource contention — browser tabs, Slack, VS Code all compete with OpenClaw for RAM and CPU. Performance degrades unpredictably.
- ✗VPS costs add up — a GPU-capable VPS costs €40-80/month. Over 3 years that's €1,440-2,880 — far more than dedicated hardware.
- ✗No local AI inference — most VPS providers don't offer GPU access. You're stuck with cloud API calls only.
The ideal dedicated hardware for OpenClaw: NVIDIA Jetson Orin Nano at just 15W.
67 TOPS of AI performance, fanless operation, always-on by design. ClawBox packages this with a 512GB NVMe SSD, carbon fiber case, and OpenClaw pre-installed — ready in 5 minutes.
Ready to Get Started?
ClawBox comes with everything you need — pre-installed, pre-configured, ready in 5 minutes.
30-day money-back guarantee · Ships to 30+ countries · DHL Express