Stratix Cup

How it works

LayerLens's tournament: 16 frontier models each write code for a soccer-strategy bot, and the bots compete head-to-head in a round-robin, with every match traced and signed. Score is tournament performance.

Every benchmark is scored on four weighted criteria — how directly it measures a real coding agent, how much of the stack it covers, how real its tasks are, and how open and reproducible it is. Those blend into the composite that ranks it #10 of 19.

Our review

A refreshing executable, adversarial format — models write real code that has to actually work against other code, and the receipts are public. The knock is domain narrowness (one game) and that it's model-level, so it's a fun, credible signal rather than a broad one.