Op
gpt-5.2
Benchmark & Insights
OpenAI OpenAI API
Updated Oct 4, 2026 All models
Sample size
232 runs
in window
Accuracy
64.4%
reference match · 233d
Confidence
77%
over 233 runs
Window end
Oct 4, 2026
most recent run
Input price
$1.75/MTok
prompt tokens
Output price
$14.00/MTok
completion tokens
Model insights
- 01 Leans high: it called 32 "MEDIUM_SAFE" days "unsafe", while missing only 4 "unsafe" days.
- 02 Hit its low point in April (50%), a calm month.
- 03 Expensive at $1.75/$14 for a model that over-warns this much.
Recent forecasts