RESCUE-BENCH - Support Strategy Prediction: leaderboard
RESCUE-BENCH: choosing the support strategy to use, ranked against the reference strategy. MRR, higher is better. Source: arxiv.org/abs/2609.09657.
Metric: MRR (%). Source: arxiv.org. Saturation forecast: Around January 2028. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V4 Pro | 52 | #112 |
| 2 | GPT-4o | 49.36 | #357 |
| 3 | Kimi K2.5 | 47.66 | #152 |
| 4 | DeepSeek V3.2 | 46.17 | #210 |
| 5 | Qwen 3 Max | 45.76 | #213 |
| 6 | Qwen 3.5 Plus | 44.98 | #133 |
| 7 | DeepSeek V4 Flash | 44.13 | #155 |
| 8 | MiniMax-M2.5 | 42.89 | #316 |
| 9 | DeepSeek R1 | 39.71 | #262 |
| 10 | Qwen Plus | 36.95 | #393 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=rescue-bench-support-strategy-prediction · How It Works · Data refreshed daily, snapshot 2026-11-05.