A privacy-first, serverless Cloudflare Worker tool for batch testing and comparing large language model (LLM) APIs.
- Privacy First: All configurations and histories are stored entirely in your browser (
localStorage). No APIs are ever captured by backend logs. - Batch Testing: Input multiple models continuously.
- Accurate Benchmarking: Tests both
Time To First Token (TTFT)and accurate Output Speed (tok/s) using an advanced heuristic. - Provider Ready: 10+ built-in presets (OpenAI, Anthropic, Gemini, DeepSeek, Alaya, Kimi, etc).
- Exporting: Save your LLM API rankings directly as a CSV.
Cloudflare Workers is the easiest platform to setup LLM-API-SpeedTest:
# clone this repository
git clone https://github.com/PoweredByPassion/LLM-API-SpeedTest.git
cd LLM-API-SpeedTest
# install Cloudflare Workers (cli) aka Wrangler
npm i
# run locally for development or testing
npm start
# deploy the worker
npm run deployAlternatively, directly string copy worker.js to the Cloudflare Dashboard in Edit Code mode.
Enter your base URL and API secret, click Fetch Models, and instantly have your entire model list dumped into your test suite.