ClinConsensus - Lay-Facing: leaderboard
ClinConsensus (Zheng et al., 2026) grades open-ended answers to expert-curated Chinese clinical cases against 30 physician-written binary criteria per case, judged by GPT-5.1; CACS@10 gives no credit below 10 met criteria and graded credit above, in percent, higher is better. This board covers the 1,004 lay-facing cases, whose stated user is a patient, relative, friend or member of the public; case construction dropped cases that DeepSeek V3, GPT-5 or Gemini 2.5 Pro answered with more than half the criteria met.
Metric: CACS@10 (%; GPT-5.1 rubric judge, lay-facing cases). Source: arxiv.org. Saturation forecast: Around October 2028. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 (High) | 33.9 | #172 (GPT-5.2) |
| 2 | Qwen 3.5 Plus | 26.1 | #227 |
| 3 | Claude Opus 4.6 | 25.2 | #100 |
| 4 | Kimi K2.5 | 23 | #252 |
| 5 | DeepSeek V3.2 | 22.9 | #403 |
| 6 | Qwen 3 Max (2026-01-23) | 22.3 | #288 |
| 7 | Gemini 3.1 Pro (Preview) | 22 | #91 |
| 8 | Grok 4.1 Fast | 19.9 | #426 |
| 9 | Seed 2.0 Pro | 19.9 | #209 |
| 10 | GLM-5 | 19.5 | #256 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=clinconsensus-lay-facing · How It Works · Data refreshed daily, snapshot 2026-11-12.