PerSo - Behavioral: leaderboard
Table 5; 200 ChangeMyView threads; behavioral pair choice or 400 four-option cognitive instances per dimension; temperature 0, top-k 1; synthetic cognitive labels generated by Gemini 2.5 Pro; not persuasion win rate. Metric: Accuracy (%). Higher is better. Source: https://arxiv.org/html/2610.20548v1
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around 2029. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 397B A17B (Non-reasoning) | 76 | #144 (Qwen 3.5 397B A17B) |
| 2 | Qwen 3.5 35B A3B (Non-reasoning) | 75 | #250 (Qwen 3.5 35B A3B) |
| 3 | Qwen 3.5 397B A17B (Thinking) | 73 | #144 (Qwen 3.5 397B A17B) |
| 4 | Qwen 3.5 35B A3B (Thinking) | 73 | #250 (Qwen 3.5 35B A3B) |
| 5 | Claude Sonnet 4.6 | 71.5 | #89 |
| 6 | Gemini 2.5 Pro | 71 | #150 |
| 7 | Gemma 3 27B | 71 | #597 |
| 8 | Claude Haiku 4.5 | 69.5 | #275 |
| 9 | DeepSeek V3.2 | 69 | #201 |
| 10 | DeepSeek V3.1 | 67.5 | #261 |
| 11 | Qwen 3.5 4B (Thinking) | 66.5 | #463 (Qwen 3.5 4B) |
| 12 | Qwen 3.5 4B (Non-reasoning) | 64 | #463 (Qwen 3.5 4B) |
| 13 | Gemma 3 12B | 62.5 | #662 |
| 14 | DeepSeek V4 Pro | 62 | #99 |
| 15 | Gemma 3 4B | 60.5 | #1074 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #7 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=perso-behavioral · How It Works · Data refreshed daily, snapshot 2026-10-16.