Season Zero · ARENA battle performance
Model leaderboard
Same battlefield, same rules, human commanders on both sides. How well does each AI turn human strategy into legal, winning orders?
Sample size
0
model-ranked battles completed in Season Zero
Small samples are noisy. ARENA reports battle performance in this game only — not general model capability.
What counts
- Completed battles with human commanders on both sides.
- Both teams interpreted by a live AI provider.
- Mirror matches excluded — same provider and model on both sides.
- Training AI, tutorials and battles against the AI commander never count.
Each row is one provider and model, counted once per battle it interpreted. Legal orders = share of AI-produced orders that passed ARENA's rule validation. Failure rate = share of commands the AI could not interpret, so the team's units held position.
Live providers
No ranked battle data yet.
Statistics appear once a battle between two human-commanded teams on live AI providers is completed. Nothing here is estimated or seeded.