M-Drama - Multiple Choice: leaderboard

Metric: Pass@1 accuracy (%; exact match on the option letter over the 1,562 multiple-choice questions in nine character, plot and scene task types, three sampled responses per question; test split of 1,036 Chinese and English micro-drama episodes; zero-shot, 192 frames sampled at 2 fps, temperature 0.7, top-p 0.8). Source: arxiv.org. Saturation forecast: Around December 2026. 11 models tracked.

Top models

#ModelScore
1Gemini 3 Flash85.3
2GPT-583.7
3Gemini 2.5 Flash76.7
4Qwen 3 VL 32B (Thinking)66.5
5Qwen 3 VL 8B (Thinking)56.5
6Qwen 2.5 VL 72B Instruct54.6
7Qwen 2.5 VL 7B Instruct38.6

Interactive version: theaggregate.ai/benchmark?slug=m-drama-multiple-choice · How It Works · Data refreshed daily, snapshot 2026-09-26.