A live timing lane for AI models.
AI Drag Racing lets you run model responses side by side and compare time to first token, total response time, throughput, and response behavior in one browser view. It is built by Jonathan R Reed for quick hands-on checks when model speed matters as much as answer quality.
The tool is meant for practical testing, not a universal leaderboard. Your prompt, provider keys, network, selected model, and provider load all affect the result, so each run is best read as a live experiment.
Use it when you need to choose between models for a concrete workflow: support drafts, coding help, summarization, structured extraction, writing, or agent steps where slow starts are noticeable. Running the same prompt across several providers makes timing differences easier to see before you commit to deeper evals.
The page reports speed metrics beside the response itself because a fast model is not useful if the answer is incomplete, terse, or formatted poorly. The best read is both mechanical and editorial: compare the time, then inspect what each model actually produced.
AI Drag Racing keeps that comparison small on purpose: one prompt, a set of model lanes, timing numbers, and the responses you can review directly. It is a quick field check before deeper evaluation work.
Published .