Skip to content

README.md

Latest commit

 

History

History
35 lines (24 loc) · 820 Bytes

README.md

File metadata and controls

35 lines (24 loc) · 820 Bytes
Raw

AccelMark Benchmark Skill for OpenClaw

Benchmark your AI accelerator and see how you rank in the AccelMark community.

What it does

  1. Detects your GPU/NPU automatically
  2. Selects the right test for your hardware (no configuration needed)
  3. Runs a 2-5 minute benchmark
  4. Shows your speed and community ranking
  5. Optionally submits to the AccelMark leaderboard

Install

In OpenClaw:

install skill AccelMark

Or from ClawHub: clawhub.ai/skills/accelmark

Usage

Just say:

  • "benchmark my GPU"
  • "run accelmark"
  • "how fast is my machine for AI?"

Requirements

  • Python 3.10+
  • 4GB+ VRAM (CPU fallback available but slow)
  • Internet connection (for model download on first run)

First run downloads the model (~4GB for 8B Q4). Subsequent runs use cache.