Observed arrival · 2026-08-21
DurkBench, the hourly AI coding-tool speed wall
DurkBench measures paid AI coding applications such as Claude Code, Codex, and Grok Build on the same lab machine, ranking their speed and clean-round rate.
Why it surfaced
This is a useful correction to API-only model leaderboards: the site tests the subscription apps people actually use, once an hour, and keeps the raw logs behind every number. Its live board reports both wall tokens per second and whether each round finished cleanly.
DurkBench measures paid AI coding applications such as Claude Code, Codex, and Grok Build on the same lab machine, ranking their speed and clean-round rate.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue