WideResearch (BenchLM): leaderboard
Research agents gather information from multiple sources and synthesize an answer. BenchLM’s attributed table reports task performance for this browsing and research workflow. These are published results, not BenchLM ability estimates.
Metric: Task Score (%). Source: benchlm.ai. Saturation forecast: Around February 2027. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Hy4 preview | 83.9 | #45 |
| 2 | Ling 3.1 Flash | 83.3 | #75 |
| 3 | Qwen 3.8 Max | 81.9 | #34 |
| 4 | Kimi K2.6 | 80.8 | #113 |
| 5 | Ornith-1.5-397B | 80.8 | |
| 6 | Claude Opus 4.5 | 76.4 | #87 |
| 7 | Qwen 3.6 Plus | 74.3 | #124 |
| 8 | Qwen 3.5 397B A17B | 74 | #154 |
| 9 | Ling-3.0-flash | 73.6 | #215 |
| 10 | Kimi K2.5 | 72.7 | #152 |
| 11 | GLM-5 | 69.8 | #149 |
| 12 | Ornith-1.5-35B-A3B | 67.8 | |
| 13 | Qwen 3.6 35B A3B | 60.1 | #238 |
| 14 | Ornith-1.5-9B | 59.5 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=wideresearch-benchlm · How It Works · Data refreshed daily, snapshot 2026-11-05.