MEGA-Bench Task - Constrained Generation Contain Length — leaderboard
Metric: Task Score (%). Source: huggingface.co. 27 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4o Mini | 86.7 |
| 2 | GPT-4o | 80 |
| 3 | Claude 3.5 Sonnet (20241022) | 73.3 |
| 4 | Qwen 2 VL 72B | 66.7 |
| 5 | InternVL2.5-78B | 66.7 |
| 6 | Gemini 1.5 Flash (002) | 40 |
| 7 | Qwen 2 VL 2B | 33.3 |
| 8 | Aquila-VL-2B | 33.3 |
| 9 | Gemini 1.5 Pro (002) | 33.3 |
| 10 | Pixtral-12B | 33.3 |
| 11 | Qwen 2 VL 7B | 26.7 |
| 12 | InternVL2.5-2B | 26.7 |
| 13 | InternVL2-8B | 6.7 |
| 14 | MiniCPM-V-2.6 | 0 |
| 15 | Claude 3.5 Sonnet (20240620) | 0 |
Interactive version: theaggregate.ai/benchmark?slug=mega-bench-task-constrained-generation-contain-length · How the rankings work · Data refreshed daily, snapshot 2026-07-22.