GPAgentBench-2K - Management Accuracy: leaderboard
Metric: Management accuracy (%; episodes ending in the correct pathway, local treatment or specialist referral, checked deterministically; 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. Saturation forecast: Around 2031. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4.6 | 60.2 |
| 2 | DeepSeek V4 Pro | 56.7 |
| 3 | GPT-5.4 | 56.3 |
| 4 | Qwen 3.7 Max | 56.1 |
| 5 | Gemini 3.1 Pro (Preview) | 55.9 |
| 6 | Claude Opus 4.7 | 55.9 |
| 7 | MiMo-V2.5-Pro | 55.3 |
| 8 | GLM-5.1 | 54.4 |
| 9 | Lingshu-32B | 53.6 |
| 10 | DeepSeek V4 Flash | 53.2 |
| 11 | MiniMax-M2.7 | 53 |
| 12 | MedGemma-27B | 52 |
| 13 | Lingshu-7B | 44.5 |
| 14 | Kimi K2.6 | 41.2 |
Interactive version: theaggregate.ai/benchmark?slug=gpagentbench-2k-management-accuracy · How It Works · Data refreshed daily, snapshot 2026-09-26.