RUN 20260725-opencode-mini-control (design + open note)

A benchmark in which language models design roller coasters that RollerCoaster Tycoon 2 builds, tests, and rates.

Standings

design mode — models build from scratch

modelscoreintensitynausea similaritybesttokens / cost
1st openrouter_openai_gpt-5-mini no successful coaster 28k in / 8k out · $0.04

score = excitement, scaled down when similarity to a stock library design exceeds 0.5 (an exact or mirrored copy scores 0)

openrouter_openai_gpt-5-minifull detail →

Round 1 build failed trace (89, 3 rejected) →

build failed (10/10 pieces placed): finish_and_test: set_status(testing): Track is not a complete circuit [track starts at {"bank":0,"dir":2,"slope":0,"x":60,"y":60,"z":112} and currently ends at {"bank":0,"dir":1,"slope":0,"x":64,"y":58,"z":112}; a closed circuit requires these to match]

track program (10 pieces)
{
  "pieces": [
    "begin_station",
    "middle_station",
    "end_station",
    "flat",
    "flat",
    "flat",
    "flat",
    "right_turn_5",
    "right_turn_5",
    "right_turn_5"
  ],
  "ride_type": 51,
  "start": {
    "dir": 2,
    "x": 60,
    "y": 60
  }
}
park screenshot, openrouter_openai_gpt-5-mini round 1

← all runs