AgentWebBench (Centralized Retrieval) - Question Answering: leaderboard

AgentWebBench (2026) models the Agentic Web: a user agent serves information needs over 100 ClueWeb22-B websites by coordinating with content agents or tools that each cover one website, with every LLM in thinking mode. This board is answer accuracy in percent judged by an LLM on DeepResearchGym short-answer questions, under centralized retrieval over one index of all websites, the reference without access constraints; higher is better.

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 February 2027. 7 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Gemini 3 Flash (Preview)60.38#156
2GPT-5 Mini35.85#373
3Qwen 3 Next 80B A3B (Thinking)33.96#606
4Qwen 3 30B A3B 2507 (Thinking)28.3#769
5DeepSeek V3.2Thinking24.53#425
6Qwen 3 4B 2507 (Thinking)18.87#1447
7Qwen 3 14B16.98#1284

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=agentwebbench-centralized-retrieval-question-answering · How It Works · Data refreshed daily, snapshot 2026-11-22.