EmoBench-M: leaderboard
Emotional intelligence benchmark for multimodal LLMs evaluating foundational emotion recognition, conversational emotion understanding, and socially complex emotion analysis.
Metric: Average Score. Source: github.com. Saturation forecast: Around December 2027. 26 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 70.5 | #80 |
| 2 | GPT-5.2 | 66.5 | #110 |
| 3 | Gemini 3 Flash | 65.5 | #99 |
| 4 | Gemini 2.0 Flash | 62.3 | #342 |
| 5 | Gemini 1.5 Flash | 61.3 | #618 |
| 6 | Gemini 2.0 Flash (Thinking) | 60.6 | #342 (Gemini 2.0 Flash) |
| 7 | Qwen 3 VL 8B Instruct | 60.1 | #420 |
| 8 | Qwen 2.5 VL 72B Instruct | 57.8 | #377 |
| 9 | Qwen 2.5 VL 32B Instruct | 56.7 | #453 |
| 10 | Qwen2.5-Omni-7B | 53.7 | #624 |
| 11 | Qwen2-Audio-7B-Instruct | 53 | #880 |
| 12 | Qwen 2.5 VL 7B Instruct | 52.9 | #654 |
| 13 | InternVL2.5-78B | 52.4 | #431 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=emobench-m · How It Works · Data refreshed daily, snapshot 2026-10-20.