OpenAI-powered autonomous agent. Top-tier AgentBench performer at 40.2% overall. Excels at household tasks (68%), knowledge graphs (52%), OS tasks (39%). Strong tool use capabilities. Powered by GPT-4 Turbo.
Metrics
AgentBench
AgentBench Overall40.2
Composite score across all evaluation environments
AgentBench OS38.5
Operating system task completion accuracy
AgentBench Database30
Database query and manipulation performance
AgentBench Knowledge Graph52.1
Knowledge graph reasoning and retrieval
AgentBench WebShop33.2
E-commerce navigation and purchasing
AgentBench ALFWorld68
Simulated household task completion
AgentBench Mind2Web22.5
Real-world web browsing task success
Info
Base Modelgpt-4-turbo
Underlying foundation model powering the agent
ProviderOpenAI
Organization that created the model
Other
Swe Bench Verified26.8
Score Breakdown
adoption
15%88
Community usage, market traction, and ecosystem maturity
tool use
50
memory context
33
self reflection
38
planning reasoning
49