Demo environment: mock data only

Agent Reliability Lab

Results

Before-and-after numbers for the three projects. A few playground runs stay on the Playground page. These tabs read the eval files only, and every figure comes from those files.

openai/gpt-oss-120b · 50 tasks · 2026-11-13T11:42:41.572Z

ConditionStuck rateAvg stepsAvg costAvg latencyEscalated
Guards off0.0%2.80$0.00036359 ms0
Guards on0.0%2.80$0.00037349 ms0

Guards off versus guards on