chini-023-airline-gate-turnaround

Airline Gate Turnaround

25 minutes to deplane, refuel, clean, cater, and board 180 passengers. Anything later costs the airline a delay slot.

Source: Operations research, airline ground ops manuals

Prompt

Design the workflow for a single-gate aircraft turnaround at a major hub.

Functional:
- Inbound aircraft arrives. Passengers deplane via jetbridge. Baggage handlers unload cargo hold.
- Cleaning crew enters cabin. Caterers restock galley. Fueler tops off tanks. Lavatory truck services waste.
- Outbound passengers board via the same jetbridge once cleaning is done. Baggage handlers load outbound cargo.
- Pushback tug pushes the aircraft back. Gate is now free for the next inbound.

Non-functional:
- Total turnaround target: 25 minutes. Some steps run in parallel (cleaning + fueling), others must serialize (cannot board while cleaning).
- A 3x arrival surge during weather recovery must not collapse the gate (queue inbound flights or divert).
- If the fueler truck breaks down, the design must reroute to a backup truck or hold the next departure rather than send an aircraft out unfueled.
- Boarding cannot start before cleaning finishes (safety).

Return a Chinilla CanvasState modeling the parallel and serial workflow stages.

Constraints

Max components: 14
Required behaviors: queue, circuitbreaker, split
Monthly budget: $95000

Stress scenarios

Standard turnaround

baseline

Normal arrival cadence, no failures. Aircraft turn in target window.

Weather recovery surge

spike

Three-hour storm clears, 3x backlog hits the gate at once. Queue or divert.

Fueler truck dies

outage

Primary fueler breaks down mid-shift. Reroute to backup or hold departures.

Cleaning runs long

latency

Crew is short-staffed and each clean takes 4 extra minutes. Boarding must wait.

Pass criteria (overall)

Min stability score: 65
Max drop rate: 10.0%
Min delivery rate: 85.0%
Max errors: 8

Submit your run

Submissions go through the chini-bench CLI. It calls your model with your key, scores the result locally, and posts to the leaderboard. Nothing leaves your machine except the canvas it produces.

End-to-end:

pip install git+https://github.com/collapseindex/chini-bench-cli.git
export OPENROUTER_API_KEY=...

chini-bench run chini-023-airline-gate-turnaround \
  --provider openrouter --model google/gemini-2.0-flash-001 \
  --as alice

Or inspect the prompt first:

chini-bench prompt chini-023-airline-gate-turnaround

Providers: openai · anthropic · google · openrouter · ollama

Leaderboard

Rank	Submitter	Model	Score	Stability	Delivery	Design	Pass
#1	alex	google/gemini-3.1-pro-preview default reflexion	92	85.0	100.0	100.0	✗
#2	alex	anthropic/claude-sonnet-4.6 default single-shot	81	47.0	100.0	100.0	✗
#3	alex	anthropic/claude-sonnet-4.6 default reflexion	77	45.0	100.0	100.0	✗
#4	alex	x-ai/grok-4.20 default reflexion	72	48.0	61.0	100.0	✗
#5	alex	x-ai/grok-4.20 default single-shot	66	74.0	0.0	100.0	✗
#6	alex	openai/gpt-5.4 default single-shot	59	69.0	0.0	75.0	✗
#7	alex	google/gemini-3.1-pro-preview default single-shot	48	0.0	30.0	100.0	✗
#8	alex	openai/gpt-5.4 default reflexion	35	0.0	0.0	75.0	✗

Per-scenario breakdown of the top run

Scenario	Health	Drop rate	Delivered	Pass
baseline	85.0	0.0%	72	✓
weather-surge	85.0	0.0%	216	✓
fueler-breakdown	85.0	0.0%	64	✓
slow-cleaning	85.0	0.0%	64	✓

How is this scored? →