CarWashBench: leaderboard
CarWashBench — long-horizon business-simulation benchmark where agents operate a car-wash enterprise; composite operational score.
Metric: Score (%). Source: carwashbench.github.io. Saturation forecast: Around October 2027. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 72.5 | #58 |
| 2 | GLM-5 | 47.5 | #141 |
| 3 | Gemma 4 31B | 20 | #200 |
| 4 | GPT-5.5 (Medium) | 17.5 | #31 (GPT-5.5) |
| 5 | Gemini 3 Flash | 12.5 | #99 |
| 6 | Claude Opus 4.6 (Thinking) | 10 | #63 (Claude Opus 4.6) |
| 7 | Qwen 3.5 27B | 0 | #187 |
| 8 | O3 | 0 | #124 |
| 9 | MiniMax-M2.5 | 0 | #303 |
| 10 | Claude Sonnet 4.6 (Thinking) | 0 | #90 (Claude Sonnet 4.6) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=carwashbench · How It Works · Data refreshed daily, snapshot 2026-10-20.