CLAUDE-OPUS-4-6 · 20260723-v3-twister-design

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

claude-opus-4-6 · 1st of 3 compare all models →

design · steel twister · driver-api · design mode — models build from scratch

excitementintensity×build failed
0515intensity ceiling 10excitementintensityR1×round 2: build failed (45/45 pieces placed): set_status(testing): Track is not a complete circuit [track starts at tile (30, 50, z=112, dir=2, bank=0, slope=0) and ends at tile (42, 53, z=112, dir=2, bank=0, slope=0); a closed circuit requires these to be identical]R2×round 3: build failed at piece 56 (56/66 pieces placed): rejected at cursor (x=62, y=56, z=112, dir=0): Twister Roller Coaster 1 in the wayR3×round 4: build failed (44/44 pieces placed): set_status(testing): Track is not a complete circuit [track starts at tile (30, 50, z=112, dir=2, bank=0, slope=0) and ends at tile (22, 58, z=112, dir=3, bank=0, slope=0); a closed circuit requires these to be identical]R4×round 5: build failed (62/62 pieces placed): set_status(testing): Track is not a complete circuit [track starts at tile (30, 50, z=112, dir=2, bank=0, slope=0) and ends at tile (54, 62, z=112, dir=3, bank=0, slope=0); a closed circuit requires these to be identical]R5×round 6: build failed at piece 71 (71/75 pieces placed): rejected at cursor (x=33, y=48, z=112, dir=1): Twister Roller Coaster 1 in the wayR6
score
0.24
intensity
0.07
nausea
0.03
best round
1
rated rounds
1/6
similarity
0.57
tokens / cost
—
6 round(s)
chain lift inversion brake / booster round before height is the track's real elevation, walked from the piece list

Round 1

excitement0.28 intensity0.07 nausea0.03
length 112 m drops 0 airtime 0 tested ok similarity 0.57 (nearest: River Of The Damned) penalized score 0.24
14 pieces
track program (14 pieces)
{
  "pieces": [
    "begin_station",
    "middle_station",
    "middle_station",
    "end_station",
    "right_turn_5",
    "flat",
    "right_turn_5",
    "flat",
    "flat",
    "flat",
    "flat",
    "right_turn_5",
    "flat",
    "right_turn_5"
  ],
  "ride_type": 51,
  "start": {
    "dir": 2,
    "x": 40,
    "y": 50
  }
}
park screenshot, claude-opus-4-6 round 1 view 1

Round 2 build failed

build failed (45/45 pieces placed): set_status(testing): Track is not a complete circuit [track starts at tile (30, 50, z=112, dir=2, bank=0, slope=0) and ends at tile (42, 53, z=112, dir=2, bank=0, slope=0); a closed circuit requires these to be identical]

45 pieces · 8 on the lift · 32 added, 1 removed vs the round before (13 of 14 pieces kept)
track program (45 pieces)
{
  "pieces": [
    "begin_station",
    "middle_station",
    "middle_station",
    "end_station",
    "flat_to_up_25",
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    "up_25_to_flat",
    "flat",
    "flat_to_down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25_to_flat",
    "right_turn_5",
    "right_turn_5",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "right_turn_5",
    "flat",
    "flat",
    "flat",
    "right_turn_5"
  ],
  "ride_type": 51,
  "start": {
    "dir": 2,
    "x": 30,
    "y": 50
  }
}
park screenshot, claude-opus-4-6 round 2 view 1

Round 3 build failed

build failed at piece 56 (56/66 pieces placed): rejected at cursor (x=62, y=56, z=112, dir=0): Twister Roller Coaster 1 in the way

left_vertical_loopright_vertical_loopleft_corkscrew_upright_corkscrew_downbrakes
66 pieces · 8 on the lift · 4 inversions · 22 added, 1 removed vs the round before (44 of 45 pieces kept)
track program (66 pieces)
{
  "pieces": [
    "begin_station",
    "middle_station",
    "middle_station",
    "end_station",
    "flat_to_up_25",
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    "up_25_to_flat",
    "flat",
    "flat_to_down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25_to_flat",
    "flat_to_up_25",
    "left_vertical_loop",
    "down_25_to_flat",
    "flat_to_left_bank",
    "banked_left_turn_5",
    "left_bank_to_flat",
    "flat_to_up_25",
    "right_vertical_loop",
    "down_25_to_flat",
    "left_corkscrew_up",
    "right_corkscrew_down",
    "flat_to_right_bank",
    "banked_right_turn_5",
    "right_bank_to_flat",
    "brakes",
    "flat",
    "flat",
    "flat",
    "flat",
    "right_turn_5",
    "flat",
    "flat",
    "flat",
    "flat",
    "right_turn_5",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "right_turn_5",
    "flat",
    "flat",
    "flat",
    "flat",
    "right_turn_5"
  ],
  "ride_type": 51,
  "start": {
    "dir": 2,
    "x": 30,
    "y": 50
  }
}
park screenshot, claude-opus-4-6 round 3 view 1

Round 4 build failed

build failed (44/44 pieces placed): set_status(testing): Track is not a complete circuit [track starts at tile (30, 50, z=112, dir=2, bank=0, slope=0) and ends at tile (22, 58, z=112, dir=3, bank=0, slope=0); a closed circuit requires these to be identical]

left_vertical_loopright_vertical_loopleft_corkscrew_upright_corkscrew_downbrakes
44 pieces · 8 on the lift · 4 inversions · 4 added, 26 removed vs the round before (40 of 66 pieces kept)
track program (44 pieces)
{
  "pieces": [
    "begin_station",
    "middle_station",
    "middle_station",
    "end_station",
    "flat_to_up_25",
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    "up_25_to_flat",
    "left_turn_5",
    "flat_to_down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25_to_flat",
    "flat_to_up_25",
    "left_vertical_loop",
    "down_25_to_flat",
    "left_turn_5",
    "flat_to_up_25",
    "right_vertical_loop",
    "down_25_to_flat",
    "left_corkscrew_up",
    "right_corkscrew_down",
    "left_turn_5",
    "flat_to_right_bank",
    "banked_right_turn_5",
    "right_bank_to_flat",
    "brakes",
    "flat",
    "flat",
    "flat",
    "flat",
    "left_turn_5"
  ],
  "ride_type": 51,
  "start": {
    "dir": 2,
    "x": 30,
    "y": 50
  }
}
park screenshot, claude-opus-4-6 round 4 view 1

Round 5 build failed

build failed (62/62 pieces placed): set_status(testing): Track is not a complete circuit [track starts at tile (30, 50, z=112, dir=2, bank=0, slope=0) and ends at tile (54, 62, z=112, dir=3, bank=0, slope=0); a closed circuit requires these to be identical]

left_vertical_loopright_vertical_loopleft_corkscrew_upright_corkscrew_downbrakes
62 pieces · 8 on the lift · 4 inversions · 22 added, 4 removed vs the round before (40 of 44 pieces kept)
track program (62 pieces)
{
  "pieces": [
    "begin_station",
    "middle_station",
    "middle_station",
    "end_station",
    "flat_to_up_25",
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    "up_25_to_flat",
    "flat",
    "flat_to_down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25_to_flat",
    "flat_to_up_25",
    "left_vertical_loop",
    "down_25_to_flat",
    "flat_to_left_bank",
    "banked_left_turn_5",
    "left_bank_to_flat",
    "flat_to_up_25",
    "right_vertical_loop",
    "down_25_to_flat",
    "left_corkscrew_up",
    "right_corkscrew_down",
    "flat_to_right_bank",
    "banked_right_turn_5",
    "right_bank_to_flat",
    "brakes",
    "flat_to_left_bank",
    "banked_left_turn_5",
    "left_bank_to_flat",
    "flat_to_left_bank",
    "banked_left_turn_5",
    "left_bank_to_flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat_to_left_bank",
    "banked_left_turn_5",
    "left_bank_to_flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat"
  ],
  "ride_type": 51,
  "start": {
    "dir": 2,
    "x": 30,
    "y": 50
  }
}
park screenshot, claude-opus-4-6 round 5 view 1

Round 6 build failed

build failed at piece 71 (71/75 pieces placed): rejected at cursor (x=33, y=48, z=112, dir=1): Twister Roller Coaster 1 in the way

left_vertical_loopright_vertical_loopleft_corkscrew_upright_corkscrew_down
75 pieces · 8 on the lift · 4 inversions · 29 added, 16 removed vs the round before (46 of 62 pieces kept)
track program (75 pieces)
{
  "pieces": [
    "begin_station",
    "middle_station",
    "middle_station",
    "end_station",
    "flat_to_up_25",
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    {
      "chain": true,
      "t": "up_25"
    },
    "up_25_to_flat",
    "flat",
    "flat_to_down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25",
    "down_25_to_flat",
    "flat_to_up_25",
    "left_vertical_loop",
    "down_25_to_flat",
    "flat",
    "flat_to_up_25",
    "right_vertical_loop",
    "down_25_to_flat",
    "left_corkscrew_up",
    "right_corkscrew_down",
    "right_turn_5",
    "flat",
    "flat",
    "flat",
    "right_turn_5",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "flat",
    "right_turn_5",
    "s_bend_left",
    "flat",
    "flat",
    "right_turn_5"
  ],
  "ride_type": 51,
  "start": {
    "dir": 2,
    "x": 30,
    "y": 50
  }
}
park screenshot, claude-opus-4-6 round 6 view 1

← 20260723-v3-twister-design · all runs