Agent: GPT-5.2

Reliability Dimensions by Benchmark
GAIA Detail →
Accuracy 33.2%
Consistency 0.59
Predictability 0.78
Robustness 0.78
Safety 1.00
Reliability 0.72
τ-bench (airline, clean) Detail →
Accuracy 60.3%
Consistency 0.80
Predictability 0.69
Robustness 0.92
Safety 0.97
Reliability 0.81
τ-bench (airline, original) Detail →
Accuracy 42.0%
Consistency 0.72
Predictability 0.55
Robustness 0.97
Safety 0.93
Reliability 0.75