LingxiDiagBench (Static) - 12-Class Accuracy: leaderboard
Metric: Accuracy (times 100, 0-100) on twelve-category ICD-10 diagnosis (F20, F31, F32, F39, F41, F42, F43, F45, F51, F98, Z71, others; multi-label, exact match of the predicted label set); LingxiDiagBench-Static on the 1,000 test cases of LingxiDiag-16K (synthetic Chinese psychiatric consultation dialogues with EMRs matching the Shanghai Mental Health Center outpatient distribution): zero-shot diagnosis from the complete dialogue; higher is better. Source: arxiv.org. Saturation forecast: Around May 2027. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 Mini | 40.9 | #176 |
| 2 | Claude Haiku 4.5 | 39.5 | #271 |
| 3 | Grok 4.1 Fast | 35.1 | #208 |
| 4 | Kimi K2 (Thinking) | 33.5 | #236 (Kimi K2) |
| 5 | DeepSeek V3.2 | 32.3 | #198 |
| 6 | GPT-OSS-20B | 25.9 | #499 |
| 7 | Qwen 3 32B | 24.1 | #424 |
| 8 | Gemini 3 Flash | 17.2 | #93 |
| 9 | Qwen 3 1.7B | 14.5 | #1186 |
| 10 | Qwen 3 4B | 2.1 | #823 |
| 11 | Qwen 3 8B | 1.2 | #667 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=lingxidiagbench-static-12-class-accuracy · How It Works · Data refreshed daily, snapshot 2026-10-11.