Google-powered autonomous agent. Strong AgentBench performer at 38.1% overall. Excels at household tasks (62%), knowledge graphs (48%), OS tasks (35%). Powered by Gemini Pro.
Metrics
AgentBench Overall38.1
Composite score across all evaluation environments
AgentBench OS35.2
Operating system task completion accuracy
AgentBench Database28.5
Database query and manipulation performance
AgentBench Knowledge Graph48
Knowledge graph reasoning and retrieval
AgentBench WebShop30.1
E-commerce navigation and purchasing
AgentBench ALFWorld62
Simulated household task completion
AgentBench Mind2Web20.3
Real-world web browsing task success
Base Modelgemini-pro
Underlying foundation model powering the agent
ProviderGoogle
Organization that created the model
Score Breakdown
adoption
15%75
Community usage, market traction, and ecosystem maturity
tool use
91
memory context
35
self reflection
68
planning reasoning
59
Compatibility(2)
Sources(1)
Developed by
Google DeepMindRelated Signals(2)
Human-AI Gap Persists: 92% vs 75% on Simple Tasks
AgentsJan 11
93GAIA benchmark reveals a persistent capability gap between humans and AI on tasks trivially easy for humans. On Level 1 (simplest tasks), humans achieve 92% while best AI achieves 75%. The gap widens at higher difficulty levels, signaling fundamental limitations in AI reasoning and tool use.
Related:
Only 16% of Enterprise Deployments Are True Agents
AgentsDec 29
91Despite the agent hype, only 16% of enterprise and 27% of startup deployments qualify as true agents. Most production architectures remain simple, built around fixed-sequence or routing-based workflows.
Related: