MEGA-Bench Task - Question Solution Solving — leaderboard

Metric: Task Score (%). Source: huggingface.co. 44 models tracked.

Top models

#ModelScore
1Gemini 2.5 Pro35.7
2Gemma 3 27B (IT)35.7
3Claude 3.5 Sonnet (20241022)35.7
4GPT-4o28.6
5InternVL3-78B28.6
6InternVL3-38B28.6
7Gemini 1.5 Pro (002)28.6
8Gemini 1.5 Flash (002)28.6
9Claude 3.5 Sonnet (20240620)28.6
10Gemini 2.0 Flash (Preview)28.6
11Qwen 2 VL 72B21.4
12InternVL3-14B21.4
13GPT-4o Mini21.4
14InternVL2.5-78B21.4
15Llama 4 Scout Base21.4

Interactive version: theaggregate.ai/benchmark?slug=mega-bench-task-question-solution-solving · How the rankings work · Data refreshed daily, snapshot 2026-07-22.