AgentWebBench (Embedding-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 embedding-selected tools, websites chosen by embedding similarity and queried with 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 embedding selection (websites chosen by embedding similarity, documents fetched with a per-website dense retrieval tool, no content agents), thinking mode on; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 7 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 40.49 | #156 | |
| 2 | GPT-5 Mini | 35.41 | #373 | |
| 3 | DeepSeek V3.2 | Thinking | 35.05 | #425 |
| 4 | Qwen 3 4B 2507 (Thinking) | 31.03 | #1447 | |
| 5 | Qwen 3 30B A3B 2507 (Thinking) | 30.97 | #769 | |
| 6 | Qwen 3 14B | 27.94 | #1284 | |
| 7 | Qwen 3 Next 80B A3B (Thinking) | 27.92 | #606 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=agentwebbench-embedding-selected-tools-web-search · How It Works · Data refreshed daily, snapshot 2026-11-22.