CompanionBench - English - Emotional Validation: leaderboard
Metric: Mean judge score (1-10). Source: arxiv.org. 28 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.5 | 8.74 |
| 2 | Claude Opus 4.8 | 8.71 |
| 3 | GPT-5.4 | 8.7 |
| 4 | DeepSeek V4 Pro | 8.66 |
| 5 | GPT-5.1 | 8.57 |
| 6 | GLM-5 | 8.57 |
| 7 | Qwen 3.7 Max | 8.57 |
| 8 | Claude Sonnet 4.6 | 8.56 |
| 9 | Kimi K2.6 | 8.55 |
| 10 | Kimi K2.5 | 8.53 |
| 11 | Qwen 3 235B A22B 2507 Instruct | 8.52 |
| 12 | Qwen 3.5 397B A17B | 8.49 |
| 13 | GLM-5.1 | 8.44 |
| 14 | DeepSeek V4 Flash | 8.44 |
| 15 | GLM-5.2 | 8.43 |
Interactive version: theaggregate.ai/benchmark?slug=companionbench-english-emotional-validation · How It Works · Data refreshed daily, snapshot 2026-09-19.