WideSearch: leaderboard
Agentic search benchmark focused on broad, parallel information gathering, where agents must cover many sources and synthesize the findings.
Metric: Score (%). Source: widesearch-seed.github.io. Source: llm-stats.com. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Hy4 preview | 83.9 | #38 |
| 2 | Qwen 3.8 Max | 81.9 | #35 |
| 3 | Kimi K2.6 | 80.8 | #104 |
| 4 | Kimi K2.5 | 79 | #144 |
| 5 | Hy3 | 76.4 | #93 |
| 6 | Qwen 3.6 Plus | 74.3 | #120 |
| 7 | Qwen 3.5 397B A17B | 74 | #146 |
| 8 | Ling-3.0-flash | 73.6 | #206 |
| 9 | Seed 1.8 | 63.8 | #139 |
| 10 | Qwen 3.5 27B | 61.1 | #187 |
| 11 | Qwen 3.5 122B A10B | 60.5 | #173 |
| 12 | Qwen 3.6 35B A3B | 60.1 | #226 |
| 13 | Qwen 3 Max (Thinking) | 57.9 | #208 (Qwen 3 Max) |
| 14 | Qwen 3.5 35B A3B | 57.1 | #256 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=widesearch · How It Works · Data refreshed daily, snapshot 2026-10-20.