MMToM-QA - Goal: leaderboard
MMToM-QA multimodal theory-of-mind benchmark — goal-inference accuracy.
Metric: Goal (%). Source: chuanyangjin.com. Saturation forecast: Estimated already saturated. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Median Human | 88.5 | |
| 2 | O3 Mini | 40.7 | #272 |
| 3 | Llama 3.1 70B | 36.3 | #592 |
| 4 | Gemini 2.0 Flash (Thinking) | 34.7 | #342 (Gemini 2.0 Flash) |
| 5 | Gemini 2.0 Flash | 33.3 | #342 |
| 6 | GPT-4o | 32.3 | #343 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mmtom-qa-goal · How It Works · Data refreshed daily, snapshot 2026-10-20.