EmoHarbor - English: leaderboard
English-language setting of EmoHarbor (Ye, Xiang, Zhang and Zong, 2026; Appendix Table 5), using the same User-as-a-Judge pipeline as the Chinese headline board: profile-driven simulated users converse with the model and a user-evaluator agent rates each dialogue on ten personalised-support dimensions. Mean of the ten 1-5 Likert ratings; higher is better.
Metric: Mean of 10 Rubric Dimensions (1-5 Likert). Source: arxiv.org. Saturation forecast: Around February 2027. 17 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 4.44 | #135 |
| 2 | GPT-5.2 | 4.38 | #172 |
| 3 | Gemini 2.5 Pro | 4.04 | #262 |
| 4 | GPT-5 | 3.83 | #163 |
| 5 | Llama 3.1 8B Instruct | 3.25 | #4277 |
| 6 | Doubao-Seed-1.6 (Thinking) | 3.23 | |
| 7 | GPT-4o (2024-11-20) | 3.2 | #864 |
| 8 | O3 Mini | 3.14 | #536 |
| 9 | Llama 3 8B Instruct | 3.04 | #4652 |
| 10 | Qwen 3 8B (Thinking) | 3.02 | #1838 (Qwen 3 8B) |
| 11 | Qwen 2.5 32B Instruct | 2.9 | #1216 |
| 12 | Qwen 2.5 7B Instruct | 2.11 | #2583 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=emoharbor-english · How It Works · Data refreshed daily, snapshot 2026-11-12.