ESG Benchmark - Level 1: leaderboard
Level 1 of the ESG Benchmark (Zhao et al., 2026): 132 atomic true/false, multiple-choice and fill-in-the-blank questions on ESG concepts, standards and investing, taken from CFA ESG-certificate practice problems and mock exams. Scored by exact-answer accuracy (%); higher is better.
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 89.39 | #159 |
| 2 | Gemini 3 Pro | 87.88 | #135 |
| 3 | GPT-5 | 87.12 | #163 |
| 4 | DeepSeek R1 | 87.12 | #467 |
| 5 | O3 | 86.36 | #224 |
| 6 | Gemini 2.5 Pro | 85.61 | #262 |
| 7 | Claude Opus 4.5 | 85.61 | #152 |
| 8 | GPT-5.2 | 84.85 | #172 |
| 9 | GPT-4.1 | 83.33 | #492 |
| 10 | GPT-4o | 82.58 | #741 |
| 11 | DeepSeek V3 | 81.82 | #705 |
| 12 | Claude Sonnet 4.5 | 80.3 | #244 |
| 13 | GPT-5.1 | 80.3 | #225 |
| 14 | Gemini 2.5 Flash | 79.55 | #452 |
| 15 | Claude Haiku 4.5 | 76.52 | #568 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=esg-benchmark-level-1 · How It Works · Data refreshed daily, snapshot 2026-11-12.