List your company

Clinical Task Execution on MedAgentBench v2 (test)

76.9AccuracyGRASP
11.27628.31345.3562.387May 28, 2026

Evaluation results

MethodLinks
GRASPModel=gpt-oss-120b2026.0576.9
SkillXModel=GPT-4.12026.0574
Batch MemoryModel=GPT-4.12026.0573.4
ExpeLModel=GPT-4.12026.0572.9
No skillsModel=GPT-4.12026.0570.5
GRASPModel=GPT-4.12026.0570.3
Evo-MedAgentModel=GPT-4.12026.0569.3
ExpeLModel=gpt-oss-120b2026.0567.8
Seq. MemoryModel=GPT-4.12026.0567.7
GRASPModel=Gemini 3.1 Flash Lite2026.0567.2
Evo-MedAgentModel=gpt-oss-120b2026.0566.9
SkillXModel=Gemini 3.1 Flash Lite2026.0565.1
SkillXModel=gpt-oss-120b2026.0564.1
Batch MemoryModel=GPT-5.4 (low)2026.0563.5
GRASPModel=GPT-5.4 (low)2026.0563.3
Evo-MedAgentModel=GPT-5.4 (low)2026.0562.5
No skillsModel=gpt-oss-120b2026.0561.1
Seq. MemoryModel=GPT-5.4 (low)2026.0560.9
ExpeLModel=GPT-5.4 (low)2026.0560.4
No skillsModel=GPT-5.4 (low)2026.0559.9
Batch MemoryModel=gpt-oss-120b2026.0559.4
SkillXModel=GPT-5.4 (low)2026.0557.8
ExpeLModel=Gemini 3.1 Flash Lite2026.0555.7
Evo-MedAgentModel=Gemini 3.1 Flash Lite2026.0554.2
Seq. MemoryModel=gpt-oss-120b2026.0553.8
Batch MemoryModel=Gemini 3.1 Flash Lite2026.0551.6
Seq. MemoryModel=Gemini 3.1 Flash Lite2026.0550.5
No skillsModel=Gemini 3.1 Flash Lite2026.0544.6
GRASPModel=DeepSeek V4 Flash2026.0530.9
Evo-MedAgentModel=DeepSeek V4 Flash2026.0525.3
ExpeLModel=DeepSeek V4 Flash2026.0522.8
No skillsModel=DeepSeek V4 Flash2026.0520.4
Batch MemoryModel=DeepSeek V4 Flash2026.0516.6
SkillXModel=DeepSeek V4 Flash2026.0514.7
Seq. MemoryModel=DeepSeek V4 Flash2026.0513.8