LongBench Pro - Easy - Non-Thinking: leaderboard
Evaluates long-context understanding across English and Chinese tasks at four difficulty levels. Non-thinking and thinking results use separate published columns; missing settings remain absent. This series reports Easy - Non-Thinking using Score (%). Source: the original paper result table.
Metric: Score (%). Source: arxiv.org. Saturation forecast: Around February 2027. 34 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.7 Sonnet | 68.69 | #259 |
| 2 | Claude Sonnet 4 | 68.42 | #207 |
| 3 | Gemini 2.5 Flash | 66.55 | #255 |
| 4 | DeepSeek V3 (0324) | 65.26 | #349 |
| 5 | Qwen 3 Next 80B A3B Instruct | 65.25 | #381 |
| 6 | Kimi K2 0905 | 64.92 | #292 |
| 7 | DeepSeek V3.1 | 63.61 | #275 |
| 8 | DeepSeek V3.2 | 62.12 | #210 |
| 9 | Qwen 3 235B A22B 2507 Instruct | 61.76 | #315 |
| 10 | GPT-4o | 59.38 | #357 |
| 11 | GLM-4.6 | 58.82 | #265 |
| 12 | Qwen 3 30B A3B 2507 Instruct | 55.89 | #492 |
| 13 | GLM-4.5 | 55.68 | #283 |
| 14 | Llama 3.1 405B Instruct | 55.45 | #470 |
| 15 | Qwen 2.5 72B Instruct | 53.48 | #469 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=longbench-pro-easy-non-thinking · How It Works · Data refreshed daily, snapshot 2026-11-05.