M3Exam (Multimodal Memory) - Factual Judgment: leaderboard

Metric: Exact match (0-100) on the factual-judgment (FJ) questions of 5,150 questions over 239 long multi-session user-assistant conversations with images and PDFs (15 personas); each model answers with the annotated gold supporting sessions and their images or rendered pages placed in context, no memory module: four-option multiple choice, chance 25; printed 0-1 and x100 here; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 9 models tracked.

Top models

#ModelScore
1Qwen 3.6 Plus98
2Gemini 3.1 Pro (Preview)96.66
3GPT-5.496.21
4Seed 2.0 Pro95.32
5Claude Opus 4.694.43
6GLM-5.193.32
7Kimi K2.586.67
8Grok 479.23
9Qwen 2.5 VL 7B Instruct79.04

Interactive version: theaggregate.ai/benchmark?slug=m3exam-multimodal-memory-factual-judgment · How It Works · Data refreshed daily, snapshot 2026-09-29.