Go

gemini-3.6-flash Benchmark & Insights

Google Gemini API
Updated Jul 31, 2026 All models
Sample size
168 runs
in window
Accuracy
88.1%
consensus match · 168d
Confidence
90%
over 168 runs
Window end
Jul 31, 2026
most recent run
Input price
$1.50/MTok
prompt tokens
Output price
$7.50/MTok
completion tokens
Model insights
  • 01 The newest Gemini flash is the least accurate of the flash family, and it regressed in a clear direction: twenty-one downward misses, four of them days it called "safe" against an "unsafe" consensus.
  • 02 Older and cheaper Gemini flash builds both beat it.
Notes

20% cheaper than the previous version in output tokens. Gemini 3.6 Flash Model Card

Recent forecasts