AgentWebBench (Multi-Agent Coordination) - Deep Research: 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 key point recall in percent of long research reports on DeepResearchGym queries, under multi-agent coordination, the user agent exchanging messages with content agents that search their own sites; higher is better.
Metric: Key Point Recall (%) of long reports on 331 DeepResearchGym queries covered by the corpus, LLM user agent under multi-agent coordination (the user agent plans multi-turn exchanges with website content agents, which retrieve within their own site and answer in natural language), thinking mode on; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 7 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | GPT-5 Mini | 69.39 | #364 | |
| 2 | DeepSeek V3.2 | Thinking | 64.95 | #419 |
| 3 | Gemini 3 Flash (Preview) | 62.45 | #152 | |
| 4 | Qwen 3 Next 80B A3B (Thinking) | 57.28 | #594 | |
| 5 | Qwen 3 30B A3B 2507 (Thinking) | 52.03 | #764 | |
| 6 | Qwen 3 14B | 49.85 | #1269 | |
| 7 | Qwen 3 4B 2507 (Thinking) | 47.54 | #1433 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=agentwebbench-multi-agent-coordination-deep-research · How It Works · Data refreshed daily, snapshot 2026-11-21.