Local AI speed test
Let your AI agent
benchmark your machine.
Send one line to Claude Code, Codex, Cursor or any agent that can run commands. It measures decode and long-read prefill on your own model server. With your permission, results are uploaded to ModelDock for review and a public share card.
The optional machine name appears on your card. Your agent asks separately for local GPU use and upload/share permission. Uploads include synthetic test prompts, outputs and traces, timings, and hardware/model/runtime details—not your conversations or API keys. Submissions await review before joining the curated benchmark dataset or rankings.
