WuYuEval - Expert Module - Elo: leaderboard
Elo rating from judged pairwise comparisons of model answers on WuYuEval's 247 open-ended expert solid-waste-management questions, with about 170-190 matches per model (WuYuEval, arXiv 2608.07529v1, Appendix Table 13). Relative to this 33-configuration pool; higher is better.
Metric: Elo rating from pairwise judged comparisons. Source: arxiv.org. Saturation forecast: Around January 2027. 33 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.5 | 2022.8 | #87 |
| 2 | Kimi K2 (Thinking) | 1954.97 | #252 (Kimi K2) |
| 3 | GPT-5.2 | 1904.58 | #115 |
| 4 | DeepSeek V3.2 (Non-reasoning) | 1788.91 | #210 (DeepSeek V3.2) |
| 5 | GLM-4.6 (Non-reasoning) | 1755.15 | #265 (GLM-4.6) |
| 6 | Qwen 3 235B A22B Instruct | 1723.93 | #336 |
| 7 | Qwen 3 235B A22B (Thinking) | 1723.74 | #323 (Qwen 3 235B A22B) |
| 8 | Gemini 3 Pro (Preview) | 1719.25 | #71 |
| 9 | DeepSeek V3.2 (Thinking) | 1703.23 | #210 (DeepSeek V3.2) |
| 10 | Qwen 3 Max | 1608.84 | #213 |
| 11 | GPT-OSS-20B | 1512.59 | #523 |
| 12 | Gemma 3 12B (IT) | 1498.67 | #672 |
| 13 | Gemma 3 27B (IT) | 1458.88 | #528 |
| 14 | GLM-4.6 (Thinking) | 1423.8 | #265 (GLM-4.6) |
| 15 | Qwen 3 32B (Thinking) | 1399.3 | #449 (Qwen 3 32B) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=wuyueval-expert-module-elo · How It Works · Data refreshed daily, snapshot 2026-11-05.