LLM Stats (DROP): leaderboard
Reading comprehension requiring discrete reasoning — addition, subtraction, counting, sorting, or set operations — over numbers and entities in paragraphs.
Metric: Score (%). Source: llm-stats.com. Saturation forecast: Estimated already saturated. 29 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V3 | 91.6 | #321 |
| 2 | Claude 3.5 Sonnet | 87.1 | #350 |
| 3 | MiMo-V2.5-Pro | 86.3 | #158 |
| 4 | GPT-4 Turbo | 86 | #404 |
| 5 | Nova Pro | 85.4 | #569 |
| 6 | Llama 3.1 405B Instruct | 84.8 | #457 |
| 7 | GPT-4o | 83.4 | #343 |
| 8 | Claude 3 Opus | 83.1 | #571 |
| 9 | Claude 3.5 Haiku | 83.1 | #568 |
| 10 | GPT-4 | 80.9 | #438 |
| 11 | Nova Lite | 80.2 | #754 |
| 12 | GPT-4o Mini | 79.7 | #599 |
| 13 | Llama 3.1 70B Instruct | 79.6 | #561 |
| 14 | Nova Micro | 79.3 | #898 |
| 15 | LongCat-Flash-Chat | 79.06 | #331 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=llm-stats-drop · How It Works · Data refreshed daily, snapshot 2026-10-20.