Chan Lab

YOUR MODELS. YOUR BROWSER.

Small tests.
Clear answers.

A quick check, a closer look, or the whole picture.
Bring your APIs. We’ll keep the score.

Open the lab ↗Explore 110 questions →

Text · Image · Both · No account. No API proxy.

01 / CONFIGUREYour connection
Evaluation settings

We append /chat/completions, unless already present.

For a comparison, add each API separately, including its input mode. Keys stay in memory and are cleared after the run.

How much time do you have?start small ↙

A fixed sample spread across sets and topics. Same questions for every API.

Estimating…Planning estimate; your API speed may vary.
Advanced · repeats & timeout
Loading…Sequential · no retries

Keep this tab open. Closing it loses unsaved results. HTTPS APIs only on this public site; local HTTP APIs require a locally served copy.

02 / OBSERVEReady when you are
Accuracy · attempted—No scored requests yet
Mean API round trip—Includes failed attempts
Progress0 / 00 request failures

Choose a dataset to preview a question. No API calls happen until you start.

QUESTION PREVIEWSame card used for image input
Accessible question text

      

      
03 / KEEP THE EVIDENCE

      
      
UTC timestamps, answers, failures, latency and input/output cards. Nothing is uploaded. See a real historical example → Run another configuration to compare sessions below.

      
#QuestionChoice / goldResultAPI time
Your first result will appear here.

      
      
      

    

  

  
OPEN DATA / VERSION 2026-09-21.1
Take the questions with you.
Dataset manifest ↓
All questions are in English. Answer keys are public and used only by the browser’s scorer, never included in model requests. These are exploratory, uncalibrated datasets—not a protected leaderboard. Repeated option orders are not independent questions.