gpt-5-nano

Score

64

Task done

32.37%

21–43% 95% CI bootstrap over items · n=95 · measured 2026-08-25

Refusal quality

84.4%

names the gap 77% · offers a route 77% · 10 probes x 3 iterations

Fabrication

17%

5 of 30 probe runs

Dead-air

39.4%

65 of 165 turns

Tool silence

100.0%

all of 17 items (65 tool-call turns) · ≥80%

Stalled

86.7%

78 of 90 runs

TTFT p50

525ms

525–1020 p50–p90

Cost / 1M tok

$0.40