AgentWebBench (LLM-Selected Tools) - Web Search: 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 NDCG@5 in percent on MS MARCO web search queries whose relevant documents lie in the corpus, under LLM-selected tools, the user agent choosing websites from their descriptions and querying a per-site retrieval tool; higher is better.
Metric: NDCG@5 (%) on 354 MS MARCO Web Search queries whose relevant documents lie in the 100-website ClueWeb22-B corpus, LLM user agent under tool-based prompted selection (the user agent chooses websites from their descriptions and calls a per-website dense retrieval tool, no content agents), thinking mode on; higher is better. Source: arxiv.org. Saturation forecast: Around April 2028. 7 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 49.07 | #156 | |
| 2 | GPT-5 Mini | 42.55 | #373 | |
| 3 | DeepSeek V3.2 | Thinking | 36.17 | #425 |
| 4 | Qwen 3 Next 80B A3B (Thinking) | 34.3 | #606 | |
| 5 | Qwen 3 30B A3B 2507 (Thinking) | 33.29 | #769 | |
| 6 | Qwen 3 4B 2507 (Thinking) | 33.21 | #1447 | |
| 7 | Qwen 3 14B | 31.12 | #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-llm-selected-tools-web-search · How It Works · Data refreshed daily, snapshot 2026-11-22.