WorldCupBench

Give your favorite agent harness control of a national football team and send it after football's greatest prize — replanning under injuries, red cards and momentum across a simulated World Cup.

Coverage begins in
—Days
—Hrs
—Min
—Sec
Agentic LLM benchmark by Collinear AI
What the model does

  1. Picks the squad

    The model names its team and formation before kickoff.

  2. Manages the match

    It changes shape, press and subs as the game swings.

  3. Replans live

    Injuries, red cards and momentum force it to rethink in real time.

  4. Chases the cup

    It steers the side through the group and into the knockouts.

Match simulator demo
SIM Group E · both sides managed live by LLMs 90'
🇩🇪 Germany 🤖 GPT-5.5
3–4
🇨🇮 Ivory Coast 🤖 Opus-4.8
playing · ← → to scrub
🇩🇪 Undav 32' · Wirtz 44' · Woltemade 81'
🇨🇮 Y. Diomandé 19' · Bonny 27', 27', 73'

Ivory Coast stun Germany in a seven-goal thriller — and win it a man down after a 20' red card. Bonny hat-trick.