LongBench Pro - English - Non-Thinking: leaderboard
Metric: Score (%). Source: arxiv.org. 34 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4 | 57.14 |
| 2 | Gemini 2.5 Flash | 55.29 |
| 3 | Qwen 3 235B A22B 2507 Instruct | 52.22 |
| 4 | Claude 3.7 Sonnet | 51.89 |
| 5 | DeepSeek V3 (0324) | 51.62 |
| 6 | DeepSeek V3.2 | 50.61 |
| 7 | Qwen 3 Next 80B A3B Instruct | 50.39 |
| 8 | DeepSeek V3.1 | 50.35 |
| 9 | Kimi K2 0905 | 49.9 |
| 10 | GPT-4o | 47.67 |
| 11 | GLM-4.6 | 45.64 |
| 12 | Qwen 3 30B A3B 2507 Instruct | 43.17 |
| 13 | GLM-4.5 | 43.05 |
| 14 | Llama 3.1 405B Instruct | 42.03 |
| 15 | MiniMax-Text-01 | 40.21 |
Interactive version: theaggregate.ai/benchmark?slug=longbench-pro-english-non-thinking · How It Works · Data refreshed daily, snapshot 2026-09-19.