SGI-Bench Wet Experiment: leaderboard
SGI-Bench subscore for wet-lab experimental planning and protocol reasoning across scientific domains.
Metric: Wet Experiment Score. Source: github.com. Saturation forecast: Not forecast. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 | 36.63 | #248 |
| 2 | Qwen 3 Max | 33.62 | #208 |
| 3 | Gemini 3 Pro | 32.45 | #80 |
| 4 | GPT-4o | 31.31 | #343 |
| 5 | Qwen 3 VL 235B A22B | 30.3 | #238 |
| 6 | Claude Sonnet 4.5 | 30.15 | #142 |
| 7 | O3 | 30.04 | #124 |
| 8 | Intern-S1 | 29.02 | #285 |
| 9 | Grok 4 | 29.01 | #181 |
| 10 | O4 Mini | 28.86 | #180 |
| 11 | Claude Opus 4.1 | 25.38 | #140 |
| 12 | GPT-5.1 | 22.77 | #135 |
| 13 | Gemini 2.5 Pro | 22.05 | #152 |
| 14 | Llama 4 Scout | 21.66 | #656 |
| 15 | Gemini 2.5 Flash | 18.55 | #245 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=sgi-bench-wet-experiment · How It Works · Data refreshed daily, snapshot 2026-10-20.