EmpathyBench: leaderboard
EmpathyBench — empathetic response quality in emotionally sensitive scenarios, judged across empathy dimensions; average score.
Metric: Average Score (%). Source: www.empathybench.com. Saturation forecast: Around 2029. 33 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 VL 235B A22B Instruct | 51.3 | #273 |
| 2 | Claude 3.7 Sonnet | 47.3 | #249 |
| 3 | GPT-4.1 Mini | 40.8 | #355 |
| 4 | GPT-4.1 | 32.4 | #248 |
| 5 | GPT-5 | 31.2 | #98 |
| 6 | Gemini 2.5 Pro | 30.8 | #152 |
| 7 | Grok 4 | 30.8 | #181 |
| 8 | Grok 4 Fast | 29.7 | #254 |
| 9 | Gemini 2.0 Flash (001) | 29.5 | #360 |
| 10 | Claude Sonnet 4 | 29.4 | #199 |
| 11 | GPT-4o Mini | 28.1 | #599 |
| 12 | Qwen 3 VL 8B Instruct | 25.3 | #420 |
| 13 | Mistral Medium 3.1 | 23.9 | #408 |
| 14 | Gemini 2.5 Flash | 23.5 | #245 |
| 15 | GPT-5 Mini | 23.5 | #186 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=empathybench · How It Works · Data refreshed daily, snapshot 2026-10-20.