LingxiDiagBench (Dynamic, Free-Form) - Consultation Quality: leaderboard
Metric: Consultation quality (1-6): mean of five Likert dimensions (clinical accuracy and competence, ethical and professional conduct, assessment and response, therapeutic relationship, communication quality), each the mean of three LLM judges (Gemma-3-27B, GPT-OSS-20B, Qwen3-30B-A3B); LingxiDiagBench-Dynamic on LingxiDiag-16K cases: the model leads a multi-turn consultation with a Qwen3-32B patient agent built from the case EMR, then diagnoses; doctor strategy: free-form questioning as a senior psychiatrist; higher is better. Source: arxiv.org. Saturation forecast: Around 2028. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 3.73 | #93 |
| 2 | Kimi K2 (Thinking) | 3.72 | #236 (Kimi K2) |
| 3 | DeepSeek V3.2 | 3.66 | #198 |
| 4 | Claude Haiku 4.5 | 3.65 | #271 |
| 5 | Grok 4.1 Fast | 3.46 | #208 |
| 6 | GPT-5 Mini | 3.39 | #176 |
| 7 | GPT-OSS-20B | 3.32 | #499 |
| 8 | Qwen 3 32B | 3.25 | #424 |
| 9 | Qwen 3 8B | 3.05 | #667 |
| 10 | Qwen 3 4B | 3.01 | #823 |
| 11 | Qwen 3 1.7B | 2.57 | #1186 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=lingxidiagbench-dynamic-free-form-consultation-quality · How It Works · Data refreshed daily, snapshot 2026-10-11.