GPAgentBench-2K - Missed Local Management Rate: leaderboard
GPAgentBench-2K (Chen et al., 2026) runs LLMs zero-shot as general practitioners in interactive consultations with a simulated patient, built from 2,428 physician-validated primary-care cases and six clinical actions. Percentage of locally manageable cases the agent referred unnecessarily; lower is better.
Metric: Missed local management rate (%; locally manageable cases the agent referred unnecessarily; lower is better; zero-shot GP agent with six clinical actions over 2,428 physician-validated primary-care cases, GPT-4o-mini patient simulator, at most 10 turns). Source: arxiv.org. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.6 | 13.4 | #158 |
| 2 | MiniMax-M2.7 | 25 | #519 |
| 3 | Claude Opus 4.7 | 27.8 | #75 |
| 4 | Qwen 3.7 Max | 29.2 | #142 |
| 5 | DeepSeek V4 Pro | 33.5 | #181 |
| 6 | MiMo-V2.5-Pro | 34.5 | #274 |
| 7 | GPT-5.4 | 38.4 | #121 |
| 8 | Gemini 3.1 Pro (Preview) | 41.5 | #91 |
| 9 | GLM-5.1 | 41.5 | #175 |
| 10 | DeepSeek V4 Flash | 45.1 | #254 |
| 11 | Kimi K2.6 | 71.8 | #173 |
Interactive version: theaggregate.ai/benchmark?slug=gpagentbench-2k-missed-local-management-rate · How It Works · Data refreshed daily, snapshot 2026-11-12.