OpenTTD-Bench: leaderboard

A model in Codex at high reasoning runs a transport company in a real-time OpenTTD game for ten game years against a rule-based AI company, and the number is its cumulative operating profit in game dollars. One duel per model on one map seed; unlike turn-based business simulations, the game clock keeps running while the agent thinks, and an agent that quits early keeps its company's end-of-game profit.

Metric: Cumulative operating profit after ten game years (USD). Source: openttd-bench.pages.dev. Saturation forecast: Around February 2028. 5 models tracked. Updated 16 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1GPT-6.1 SolHigh9912626#10
2GPT-6 AstraHigh8618784#6
3GPT-5.6 SolHigh2193526#30
4GPT-5.5High1531779#52
5GPT-5.4High-46780#128

No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.

Interactive version: theaggregate.ai/benchmark?slug=openttd-bench · How It Works · Data refreshed daily, snapshot 2026-11-16.