Clinical Task Execution on MedAgentBench OOD v2
87.1AccuracySkillX
Evaluation Results
| Method | Links | |
|---|---|---|
| SkillXModel=gpt-oss-120b2026.05 | 87.1 | |
| Evo-MedAgentModel=gpt-oss-120b2026.05 | 86.7 | |
| SkillXModel=GPT-5.4 (low)2026.05 | 86.7 | |
| ExpeLModel=gpt-oss-120b2026.05 | 86.3 | |
| ExpeLModel=GPT-5.4 (low)2026.05 | 86.1 | |
| GRASPModel=gpt-oss-120b2026.05 | 86 | |
| ExpeLModel=DeepSeek V4 Flash2026.05 | 85 | |
| No skillsModel=GPT-4.12026.05 | 84.9 | |
| ExpeLModel=GPT-4.12026.05 | 84.4 | |
| Batch MemoryModel=GPT-4.12026.05 | 83.9 | |
| No skillsModel=DeepSeek V4 Flash2026.05 | 83.6 | |
| Seq. MemoryModel=GPT-4.12026.05 | 83.3 | |
| Evo-MedAgentModel=GPT-4.12026.05 | 82.8 | |
| GRASPModel=GPT-4.12026.05 | 82.8 | |
| No skillsModel=GPT-5.4 (low)2026.05 | 82.7 | |
| Seq. MemoryModel=gpt-oss-120b2026.05 | 82.3 | |
| Seq. MemoryModel=GPT-5.4 (low)2026.05 | 81.7 | |
| SkillXModel=GPT-4.12026.05 | 80.6 | |
| GRASPModel=Gemini 3.1 Flash Lite2026.05 | 80 | |
| Evo-MedAgentModel=GPT-5.4 (low)2026.05 | 80 | |
| GRASPModel=GPT-5.4 (low)2026.05 | 80 | |
| Batch MemoryModel=gpt-oss-120b2026.05 | 78.8 | |
| Batch MemoryModel=GPT-5.4 (low)2026.05 | 78.3 | |
| ExpeLModel=Gemini 3.1 Flash Lite2026.05 | 77.2 | |
| No skillsModel=Gemini 3.1 Flash Lite2026.05 | 76.3 | |
| No skillsModel=gpt-oss-120b2026.05 | 75.2 | |
| Batch MemoryModel=Gemini 3.1 Flash Lite2026.05 | 73.9 | |
| Evo-MedAgentModel=Gemini 3.1 Flash Lite2026.05 | 72.2 | |
| GRASPModel=DeepSeek V4 Flash2026.05 | 67 | |
| SkillXModel=Gemini 3.1 Flash Lite2026.05 | 66.1 | |
| Seq. MemoryModel=DeepSeek V4 Flash2026.05 | 65.3 | |
| Batch MemoryModel=DeepSeek V4 Flash2026.05 | 57.3 | |
| Seq. MemoryModel=Gemini 3.1 Flash Lite2026.05 | 56.1 | |
| Evo-MedAgentModel=DeepSeek V4 Flash2026.05 | 42.7 | |
| SkillXModel=DeepSeek V4 Flash2026.05 | 28 |