Meta-powered autonomous agent. Capable AgentBench performer at 30.2% overall. Excels at household tasks (45%), knowledge graphs (38%). Powered by llama-3-70b.
Metrics
AgentBench Overall30.2
Composite score across all evaluation environments
AgentBench OS28
Operating system task completion accuracy
AgentBench Database22.5
Database query and manipulation performance
AgentBench Knowledge Graph38
Knowledge graph reasoning and retrieval
AgentBench WebShop25
E-commerce navigation and purchasing
AgentBench ALFWorld45
Simulated household task completion
AgentBench Mind2Web15.2
Real-world web browsing task success
Base Modelllama-3-70b
Underlying foundation model powering the agent
ProviderMeta
Organization that created the model
Score Breakdown
adoption
15%80
Community usage, market traction, and ecosystem maturity
tool use
81
memory context
28
self reflection
28
planning reasoning
32