AgentWebBench (Centralized Retrieval) - Deep Research: leaderboard
Metric: Key Point Recall (%) of long reports on 331 DeepResearchGym queries covered by the corpus, LLM user agent under centralized retrieval (the agent retrieves from one index over all 100 websites, the no-access-constraint reference), thinking mode on; higher is better. Source: arxiv.org. Saturation forecast: Around September 2027. 7 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | DeepSeek V3.2 (Thinking) | 66.54 |
| 2 | Gemini 3 Flash (Preview) | 63.58 |
| 3 | GPT-5 Mini | 60.97 |
| 4 | Qwen 3 Next 80B A3B (Thinking) | 59.11 |
| 5 | Qwen 3 30B A3B 2507 (Thinking) | 54.99 |
| 6 | Qwen 3 14B | 54.4 |
| 7 | Qwen 3 4B 2507 (Thinking) | 52.74 |
Interactive version: theaggregate.ai/benchmark?slug=agentwebbench-centralized-retrieval-deep-research · How It Works · Data refreshed daily, snapshot 2026-10-07.