AgenticRAGTracer (Comparison) - 2-Hop: leaderboard
Metric: Exact match (%) on AgenticRAGTracer's 2-hop parallel comparison questions (facts about several entities gathered, then compared); every model runs as a ReAct agent in the Qwen-Agent framework, retrieving from a Wikipedia corpus (the FlashRAG collection); questions are kept only when they need retrieval, every supporting document is necessary and three annotators approve them; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Grok 4 | 80.7 | #169 |
| 2 | Qwen 2.5 72B Instruct | 78.8 | #436 |
| 3 | O4 Mini | 77.7 | #172 |
| 4 | GPT-5 | 76.2 | #91 |
| 5 | Grok 3 | 75.6 | #296 |
| 6 | DeepSeek R1 | 74.1 | #245 |
| 7 | Qwen 2.5 32B Instruct | 70.7 | #491 |
| 8 | GPT-4o | 69.2 | #333 |
| 9 | DeepSeek V3 | 66 | #312 |
| 10 | Qwen 2.5 7B Instruct | 52.4 | #846 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=agenticragtracer-comparison-2-hop · How It Works · Data refreshed daily, snapshot 2026-10-11.