MLB: leaderboard
MLB is a scenario-driven Medical LLM Benchmark for clinical applications. It combines foundational medical knowledge and scenario-based reasoning across five dimensions: medical knowledge, safety and ethics, medical-record understanding, smart services, and smart healthcare. The benchmark integrates 22 datasets, including 17 newly curated datasets from Chinese clinical sources, covering 64 clinica
Metric: Average (self-reported). Source: benchmarklist.com. Saturation forecast: Around December 2026. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K2 | 77.29 | #243 |
| 2 | Claude Sonnet 4 | 74.53 | #199 |
| 3 | GPT-5 | 73.87 | #98 |
| 4 | DeepSeek V3 | 70.5 | #321 |
| 5 | Qwen 3 235B A22B | 65.89 | #313 |
| 6 | GPT-4o | 65.45 | #343 |
| 7 | Llama 3.3 70B | 54.02 | #579 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mlb · How It Works · Data refreshed daily, snapshot 2026-10-20.