Psi-Bench - Psychological Consultation Personalization: leaderboard

Metric: Personalized response score (how well responses are tailored to the hidden client profile), psychotherapy queries (1-9; DeepSeek-v3.2 judge, which also plays the persona-profiled client; three rounds after the client's opening message, client profile hidden from the model). Source: arxiv.org. Saturation forecast: Around December 2026. 10 models tracked.

Top models

#ModelScore
1DeepSeek V4 Pro7.66
2Gemini 3.1 Pro (Preview)7.49
3Gemini 3 Flash7.37
4Qwen 3 Next 80B A3B7.26
5GPT-5.16.9
6DeepSeek V3.26.8
7Grok 4 Fast6.44
8Qwen 3 32B6.38
9GPT-5 Mini5.98
10Qwen 3 8B5.18

Interactive version: theaggregate.ai/benchmark?slug=psi-bench-psychological-consultation-personalization · How It Works · Data refreshed daily, snapshot 2026-09-29.