Skip to content

Correct Wi-Fi spec and local-model performance in hardware docs #273

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
yalexx wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Correct Wi-Fi spec and local-model performance in hardware docs #273

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
19 changes: 13 additions & 6 deletions docs-site/hardware/clawbox-connect.mdx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,9 +19,9 @@ The entry-tier ClawBox: a compact, low-power, always-on AI assistant for your de
| GPU | 1024-core NVIDIA Ampere, 625 MHz |
| Memory | 8 GB LPDDR5, 102 GB/s bandwidth (unified CPU/GPU) |
| Storage | 512 GB NVMe SSD (PCIe Gen3 x4, ~2,100 MB/s) |
| Connectivity | Wi-Fi 6 + Bluetooth 5.0 + Gigabit Ethernet |
| Connectivity | Wi-Fi 5 (802.11ac, dual-band 2×2, up to 867 Mbps) + Bluetooth 5.0 + Gigabit Ethernet |
| Ports | USB-A, USB-C, HDMI, Ethernet, microSD |
| Power draw | 7 W idle · 15 W typical · 25 W max (20 W USB-C PSU included) |
| Power draw | 7 W idle · ~11 W typical / 19 W peak measured · 25 W max (20 W USB-C PSU included) |
| Size | 100 × 79 × 31 mm, carbon-color case (~260 g) |
| OS | Ubuntu 22.04 LTS + OpenClaw pre-installed |

Expand All @@ -35,10 +35,17 @@ The entry-tier ClawBox: a compact, low-power, always-on AI assistant for your de

| Model | Speed | Quality |
|---|---|---|
| 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 |
| Llama 3.2 1B (Q4) | ~45 tok/s | Fast |
| Gemma 2 2B (Q4) | ~26 tok/s | Good |
| Qwen2.5 3B (Q4) | ~23 tok/s | Excellent |
| Phi-3 Mini 3.8B (Q4) | ~22 tok/s | Excellent |
| 7–8B class (Q4) | ~10 tok/s | Tight fit |

Measured on a production ClawBox with Ollama (warm runs, `num_predict=200`, Ollama's own
`eval_rate`; power and thermals from `tegrastats` over 176 samples): ~11.4 W average,
19.2 W peak, 61.8 °C peak, no throttling. Most people read at 5–8 tok/s, so the 1–4B
class generates faster than you can read. Full methodology:
[We Benchmarked a Production ClawBox](https://clawbox.com/blog/2026-07-22-jetson-orin-nano-llm-benchmark-real-clawbox-numbers).

See full requirements and tiers in [Hardware Requirements](/hardware/requirements).

Expand Down
21 changes: 13 additions & 8 deletions docs-site/hardware/requirements.mdx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -48,7 +48,7 @@ This page covers what you actually need — from bare minimum to running local A

### Power User — local AI models
- **Specs:** 4+ cores, 16 GB+ RAM, 100 GB+ SSD, 25 Mbps+, NVIDIA GPU, Linux (Ubuntu/Debian)
- **Can do:** everything above, plus local LLMs (7B–13B), local image generation, on-device speech recognition, fully offline operation, multiple browser instances
- **Can do:** everything above, plus local LLMs (1B–8B; 1–4B is the sweet spot on ClawBox Connect), local image generation, on-device speech recognition, fully offline operation, multiple browser instances
- **Can't do:** 70B+ models need more VRAM → see [ClawBox Workstation](/hardware/clawbox-workstation)
- **Examples:** **ClawBox Connect** (67 TOPS) · gaming PC with GPU · **ClawBox Workstation** (DGX Spark)

Expand All @@ -68,12 +68,17 @@ Running OpenClaw on dedicated hardware vs a shared VPS or your daily-use laptop:

| Model | Speed | Quality |
|---|---|---|
| Llama 3.1 8B (Q4) | ~3.5 tok/s | Excellent |
| Mistral 7B (Q4) | ~4.2 tok/s | Excellent |
| Qwen2.5 7B (Q4) | ~4.0 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 |
| Llama 3.2 1B (Q4) | ~45 tok/s | Fast |
| Gemma 2 2B (Q4) | ~26 tok/s | Good |
| Qwen2.5 3B (Q4) | ~23 tok/s | Excellent |
| Phi-3 Mini 3.8B (Q4) | ~22 tok/s | Excellent |
| 7–8B class (Q4) | ~10 tok/s | Tight fit |

Measured on a production ClawBox with Ollama (warm runs, `num_predict=200`, Ollama's own
`eval_rate`; power and thermals from `tegrastats` over 176 samples): ~11.4 W average,
19.2 W peak, 61.8 °C peak, no throttling. Most people read at 5–8 tok/s, so the 1–4B
class generates faster than you can read. Full methodology:
[We Benchmarked a Production ClawBox](https://clawbox.com/blog/2026-07-22-jetson-orin-nano-llm-benchmark-real-clawbox-numbers).

<Note>
For larger models, use your own cloud API key (Claude, GPT, Gemini) — see
Expand All @@ -84,6 +89,6 @@ see [ClawBox Workstation](/hardware/clawbox-workstation) (128 GB unified memory,
## Skip the DIY

<Card title="Get ClawBox — pre-configured, ready in 5 minutes" href="https://clawbox.tech" icon="cart-shopping" horizontal>
ClawBox ships with Recommended+ specs, OpenClaw pre-installed, dual-band Wi-Fi and Bluetooth.
ClawBox ships with Recommended+ specs, OpenClaw pre-installed, dual-band Wi-Fi 5 (802.11ac) and Bluetooth 5.0.
Manual DIY setup is typically 2–4+ hours (plus waiting for hardware).
</Card>
Loading