AgentWebBench (Centralized Retrieval) - Question Answering: leaderboard
Metric: Answer accuracy (%, LLM judge) on 53 DeepResearchGym short-answer questions 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 January 2027. 7 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 60.38 |
| 2 | GPT-5 Mini | 35.85 |
| 3 | Qwen 3 Next 80B A3B (Thinking) | 33.96 |
| 4 | Qwen 3 30B A3B 2507 (Thinking) | 28.3 |
| 5 | DeepSeek V3.2 (Thinking) | 24.53 |
| 6 | Qwen 3 4B 2507 (Thinking) | 18.87 |
| 7 | Qwen 3 14B | 16.98 |
Interactive version: theaggregate.ai/benchmark?slug=agentwebbench-centralized-retrieval-question-answering · How It Works · Data refreshed daily, snapshot 2026-10-07.