OpenClaw Observability
Usage and token estimation
Usage
Heartbeat Model
gpt-5.4-mini
Model dedicated for heartbeat and lightweight routine checks.
Current Active Model
gpt-5.4
Main active lane currently shown on the dashboard.
Estimated Token Usage
Daily input
~64.5k avg input tokens per tracked run
Daily output
~4.4k avg output tokens per tracked run
Context window
Up to 200k context on active lane, with heavier runs reaching 60%+ of limit
Cache efficiency
High cache reuse on repeated operational flows, often 50%+ cached on active sessions
Usage Notes
The page is now intentionally focused only on model usage and token estimation.
Token figures are based on tracked OpenClaw run patterns, not just static placeholders.
Use this as the quick-glance view for active model lane, heartbeat lane, and rough token behavior.