Go
gemini-3-flash-preview
Benchmark & Insights
Google Gemini API
Updated Oct 4, 2026 All models
Sample size
232 runs
in window
Accuracy
74.4%
reference match · 250d
Confidence
96%
over 252 runs
Window end
Oct 4, 2026
most recent run
Input price
$0.50/MTok
prompt tokens
Output price
$3.00/MTok
completion tokens
Model insights
- 01 Mid-pack and optimistic: zero false alarms, but 35 "MEDIUM_SAFE" days called "low" and 11 missed "unsafe" days.
- 02 Missed both late-September "HIGH" days in a 6-day wrong run.
- 03 At $0.50/$3 it still beats the pricier newer 3.6, 3.7 and 3.8 Flash models.
Notes
1M token context window
Recent forecasts