MultiEmo-Bench - Dominant Emotion: leaderboard
MultiEmo-Bench (2026) dominant emotion: a multimodal model rates how strongly each of eight emotions is evoked by an image, and its top-rated emotion is checked against the emotion most voted by 20 annotators, over 10,344 images. Precision in percent, with fractional credit for tied predictions; higher is better.
Metric: Dominant-emotion precision (%): the model scores each of the eight Mikels emotions 0-100 for how much people will feel it, and its top-scored emotion is checked against the image's most-voted emotion from 20 annotators (fractional credit 1/k when k emotions tie in the prediction; images with several dominant emotions accept any), over the 10,344 images of MultiEmo-Bench; higher is better. Source: arxiv.org. Saturation forecast: Around September 2027. 11 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 51.05 | #178 |
| 2 | GPT-5.1 | 48.14 | #240 |
| 3 | GPT-4o | 45.9 | #751 |
| 4 | Qwen 3 VL 4B | 44.9 | #1151 |
| 5 | Qwen 3 VL 32B | 44.35 | #622 |
| 6 | Qwen 3 VL 8B | 44.07 | #917 |
| 7 | Gemini 2.5 Flash | 43.62 | #463 |
| 8 | Gemini 3 Flash (Preview) | 43.2 | #152 |
| 9 | Claude Sonnet 4 | 35.1 | #410 |
| 10 | GPT-5 Nano | 31.62 | #906 |
| 11 | Qwen 3 VL 2B | 26.51 | #2134 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=multiemo-bench-dominant-emotion · How It Works · Data refreshed daily, snapshot 2026-11-21.