SGI-Bench: leaderboard
Scientific General Intelligence benchmark testing LLMs across deep research, idea generation, dry/wet experiments, and experimental reasoning. Evaluates scientific discovery capabilities.
Metric: SGI-Score. Source: github.com. Saturation forecast: Around 2030. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 33.83 | #80 |
| 2 | Claude Sonnet 4.5 | 32.16 | #142 |
| 3 | Qwen 3 Max | 31.97 | #208 |
| 4 | GPT-4.1 | 31.45 | #248 |
| 5 | GPT-5.2 Pro | 31.09 | #61 |
| 6 | GPT-5 | 30.84 | #98 |
| 7 | O3 | 30.68 | #124 |
| 8 | Claude Opus 4.1 | 30.42 | #140 |
| 9 | O4 Mini | 30.14 | #180 |
| 10 | GPT-5.1 | 29.31 | #135 |
| 11 | Grok 4 | 28.68 | #181 |
| 12 | Qwen 3 VL 235B A22B | 28.32 | #238 |
| 13 | Gemini 2.5 Pro | 28.17 | #152 |
| 14 | Intern-S1 | 28.1 | #285 |
| 15 | GPT-4o | 26.87 | #343 |
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 · How It Works · Data refreshed daily, snapshot 2026-10-20.