Anthropic-powered autonomous agent. Strong AgentBench performer at 35.4% overall. Excels at household tasks (55%), knowledge graphs (45%), OS tasks (33%). Powered by Claude 3 Opus.
Metrics
AgentBench
AgentBench Overall35.4
Composite score across all evaluation environments
AgentBench OS33
Operating system task completion accuracy
AgentBench Database26.2
Database query and manipulation performance
AgentBench Knowledge Graph45
Knowledge graph reasoning and retrieval
AgentBench WebShop28
E-commerce navigation and purchasing
AgentBench ALFWorld55
Simulated household task completion
AgentBench Mind2Web18.5
Real-world web browsing task success
Info
Base Modelclaude-3-opus
Underlying foundation model powering the agent
ProviderAnthropic
Organization that created the model
Score Breakdown
adoption
15%70
Community usage, market traction, and ecosystem maturity
tool use
40
memory context
32
self reflection
35
planning reasoning
38