AutoResearchBench - Wide Research: leaderboard
Metric: Intersection over union (%) between the collected paper set and the gold set of papers satisfying the query conditions on the Wide Research queries, a ReAct agent with the DeepXiv scientific-literature search tool, at most 30 turns; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 13 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 9.31 |
| 2 | GPT-5.4 | 8.12 |
| 3 | Seed 2.0 Pro | 7.87 |
| 4 | DeepSeek V3.2 | 7.7 |
| 5 | Qwen 3 Max | 6.89 |
| 6 | Claude Opus 4.6 | 6.56 |
| 7 | Kimi K2.5 | 6.23 |
| 8 | Claude Sonnet 4.6 | 5.83 |
| 9 | Qwen 3.5 397B A17B | 3.83 |
| 10 | Qwen 3.5 122B A10B | 2.76 |
| 11 | Qwen 3.5 35B A3B | 2.71 |
| 12 | MiniMax-M2.5 | 1.39 |
Interactive version: theaggregate.ai/benchmark?slug=autoresearchbench-wide-research · How It Works · Data refreshed daily, snapshot 2026-10-07.