Loong: leaderboard
Loong — long-context benchmark of multi-document question answering in financial, legal and academic settings where evidence is spread across every document.
Metric: Overall Avg Score (%). Source: github.com. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 1.5 Pro | 55.37 | #445 |
| 2 | GPT-4o | 53.47 | #343 |
| 3 | Claude 3.5 Sonnet | 48.85 | #350 |
| 4 | Claude 3 Haiku | 44.88 | #803 |
| 5 | Qwen 2 72B Instruct | 43.29 | #576 |
| 6 | GPT-4o Mini | 42.95 | #599 |
| 7 | GLM-4 9B Chat | 38.31 | #916 |
| 8 | Llama 3.1 8B Instruct | 36.31 | #1035 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=loong · How It Works · Data refreshed daily, snapshot 2026-10-20.