stanford/medagentbench

MedAgentBench: clinically-relevant tasks across 10 categories in a FHIR-compliant virtual EHR, benchmarking LLM agents on medical decision-making, planning, and execution.

← Back to Registry

Run this task

CLI:

inspect eval inspect_harbor/stanford_medagentbench --model openai/gpt-5

Python:

from inspect_ai import eval
from inspect_harbor import stanford_medagentbench

eval(stanford_medagentbench(), model="openai/gpt-5")

Dataset information

Harbor registry	stanford/medagentbench
Inspect task	stanford_medagentbench
Latest digest	sha256:c52d82b3462fb26417707682095e43f224a31f8f785eb7da615c1ab6adc20bf0
Samples	300
Paper	arxiv
Source	https://github.com/stanfordmlgroup/MedAgentBench

See Task Parameters for the parameter set shared across all Harbor tasks.