Op
gpt-5.4-mini
Benchmark & Insights
OpenAI OpenAI API
Updated Oct 4, 2026 All models
Sample size
232 runs
in window
Accuracy
67.9%
reference match · 268d
Confidence
94%
over 270 runs
Window end
Oct 4, 2026
most recent run
Input price
$0.75/MTok
prompt tokens
Output price
$4.50/MTok
completion tokens
Model insights
- 01 Gets the worst of both: it leans high, calling 28 "MEDIUM_SAFE" days "unsafe", yet still misses 7 "unsafe" days.
- 02 Fell below 50% in July.
- 03 The older, cheaper gpt-5-mini is far better.