Wiki Live Challenge - Wikipedia Fact Coverage: leaderboard
How much of a Wikipedia Good Article's factual content a deep research agent's own article on the same topic covers, over 100 Good Articles created in 2025; Wiki Live Challenge (Wang et al., 2026). Mean coverage score in percent from a Gemini-2.5-flash fact checker that compares each extracted Wikipedia fact with the generated article's best-matching statements; higher is better.
Metric: Fact coverage (%; mean consistency score of the Wikipedia article's extracted facts against the top-10 matching generated statements, Gemini-2.5-flash fact checker, 100 articles). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro Deep Research | 30.76 | |
| 2 | Perplexity Deep Research | 29.21 | |
| 3 | Gemini 3 Pro Deep Research | 28.83 | |
| 4 | OpenAI o3 Deep Research | 25.12 | |
| 5 | Doubao Deep Research | 22.97 | |
| 6 | Tongyi-DeepResearch-30B-A3B | 22.73 | |
| 7 | Qwen3-Max Deep Research | 22.22 | |
| 8 | langchain-open-deep-research-gpt-5 | 20.96 | |
| 9 | Grok Deep Search | 20.73 | |
| 10 | langchain-open-deep-research | 7.08 |
Interactive version: theaggregate.ai/benchmark?slug=wiki-live-challenge-wikipedia-fact-coverage · How It Works · Data refreshed daily, snapshot 2026-11-12.