RubberDuckBench: leaderboard
RubberDuckBench (Bryn Mawr) — explaining and debugging code the way a rubber-duck partner would; performance score.
Metric: Performance (%). Source: cs.brynmawr.edu. Saturation forecast: Around June 2027. 20 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Grok 4 | 69.29 | #181 |
| 2 | Claude Opus 4 | 68.53 | #163 |
| 3 | GPT-5 | 67.8 | #98 |
| 4 | Claude Opus 4.1 | 67.02 | #140 |
| 5 | O3 | 64.93 | #124 |
| 6 | Gemini 2.5 Flash | 64.3 | #245 |
| 7 | Gemini 2.5 Pro | 64.01 | #152 |
| 8 | GPT-OSS-20B | 63.63 | #518 |
| 9 | Claude Sonnet 4 | 61.66 | #199 |
| 10 | Claude 3.7 Sonnet | 61.47 | #249 |
| 11 | GPT-OSS-120B | 59.54 | #341 |
| 12 | GPT-4.1 | 59.47 | #248 |
| 13 | Llama 3.3 70B | 56.36 | #579 |
| 14 | Grok 3 | 54.74 | #304 |
| 15 | DeepSeek R1 | 54.4 | #251 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=rubberduckbench · How It Works · Data refreshed daily, snapshot 2026-10-20.