Alibaba-powered autonomous agent. Capable AgentBench performer at 25.1% overall. Excels at household tasks (38%), knowledge graphs (32%). Powered by qwen-72b.
Metrics
AgentBench
AgentBench Overall25.1
Composite score across all evaluation environments
AgentBench OS23
Operating system task completion accuracy
AgentBench Database18.5
Database query and manipulation performance
AgentBench Knowledge Graph32
Knowledge graph reasoning and retrieval
AgentBench WebShop21
E-commerce navigation and purchasing
AgentBench ALFWorld38
Simulated household task completion
AgentBench Mind2Web12.5
Real-world web browsing task success
Info
Base Modelqwen-72b
Underlying foundation model powering the agent
ProviderAlibaba
Organization that created the model
Score Breakdown
adoption
15%55
Community usage, market traction, and ecosystem maturity
tool use
30
memory context
22
self reflection
23
planning reasoning
27