AutoResearchBench - Wide Research: leaderboard
Wide Research on AutoResearchBench (2026): a ReAct agent with a scientific literature search tool must collect every paper that satisfies a set of conditions, within 30 turns. Intersection over union with the gold set, in percent; higher is better.
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 2028. 13 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 9.31 | #99 | |
| 2 | GPT-5.4 | 8.12 | #135 | |
| 3 | Seed 2.0 Pro | 7.87 | #222 | |
| 4 | DeepSeek V3.2 | 7.7 | #419 | |
| 5 | Qwen 3 Max | 6.89 | #415 | |
| 6 | Gemini 3 Flash | Non-reasoning | 6.61 | #175 |
| 7 | Claude Opus 4.6 | 6.56 | #113 | |
| 8 | Kimi K2.5 | 6.23 | #266 | |
| 9 | Claude Sonnet 4.6 | 5.83 | #173 | |
| 10 | Qwen 3.5 397B A17B | 3.83 | #264 | |
| 11 | Qwen 3.5 122B A10B | 2.76 | #346 | |
| 12 | Qwen 3.5 35B A3B | 2.71 | #492 | |
| 13 | MiniMax-M2.5 | 1.39 | #600 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=autoresearchbench-wide-research · How It Works · Data refreshed daily, snapshot 2026-11-21.