DeepResearch Bench II - Analysis: leaderboard
The analysis rubrics of DeepResearch Bench II (Li et al., 2026): the conclusions and inferences an expert review article draws, which a deep research report on the same question must reach. Share of these rubrics a report satisfies over 132 tasks, judged by Gemini-2.5-Pro; higher is better.
Metric: Rubric pass rate (%; analysis rubrics; 132 report tasks in 22 domains; share of expert-derived binary rubrics a report satisfies, Gemini-2.5-Pro judge). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro Deep Research | 51.91 | |
| 2 | o3 Deep Research | 49.85 | |
| 3 | Doubao Deep Research | 49.43 | |
| 4 | Gemini 3 Pro Deep Research | 48.94 | |
| 5 | Qwen3-Max Deep Research | 48.04 | |
| 6 | Perplexity Deep Research | 44.47 | |
| 7 | Grok Deep Search | 42.5 | |
| 8 | Tongyi-DeepResearch-30B-A3B | 35.89 |
Interactive version: theaggregate.ai/benchmark?slug=deepresearch-bench-ii-analysis · How It Works · Data refreshed daily, snapshot 2026-11-12.