AI BENCHY
Advertise here
#96

Gemini 3.8 Flash

Google Release: 2026-09-02 Tested on: 2026-09-02 20:06 google/gemini-3.8-flash::low
~500B total (~20B active)MoEClosedEstimated
(high) (medium) (low)

Summary

Gemini 3.8 Flash scores 7.5 on AI BENCHY and ranks #96. It has 10.0 reliability, a 81.8% pass rate, $0.239 total cost, and 4.83s average response time.

What makes Gemini 3.8 Flash unique: It stands out most in Trivia, where it ranks #1, while Coding is its weakest area at #16.

Model facts

Researched on 2026-09-02

Estimated
Parameters
~500B total (~20B active)
Architecture
MoE
Availability
Closed
License
-

Best estimate from public evidence; the vendor did not disclose every value. Google documents this model as based on Gemini 3.7 Flash but does not disclose architecture or parameter counts; estimates follow the related 3.7 Flash record.

Parameter source Availability source

Score

7.5

Consistency

8.5

Reliability

10.0

Total Cost (Current Price)

$0.239

Total Output Tokens

41,324

Total Input Tokens

111,773

Input Price

$0.750 / 1M

Output Price

$3.750 / 1M

Tests Correct

Wrong Tests: 6

Attempt pass rate: 81.8%

Flaky tests

4

Flaky tests had mixed outcomes across runs (at least one pass and one fail).

Response Time (avg)

4.83s

Response Time (max): 35.14s

Response Time (total): 106.18s

Wrong answer: 5 No answer: 1

Hamster playing table tennis

Prompt: Create a detailed SVG illustration of a hamster playing table tennis.

#96 Gemini 3.8 Flash

low
Cost
$0.030
Time
45.8s
Tokens
8,023 tok

Charts

Choose the first model, then click a second model to open a side-by-side page.

Total Output Tokens

Score vs Total Output Tokens

Quick Compare

Gemini 3.8 FlashlowvsGrok Build 0.1mediumGemini 3.8 FlashlowvsGemini 3.1 Flash Lite PreviewmediumGemini 3.8 FlashlowvsGPT-5.4 MinimediumGemini 3.8 FlashlowvsQwen3.7 MaxnoneGemini 3.8 FlashlowvsMercury 2.5 PreviewmediumGemini 3.8 FlashlowvsKimi K2.7 CodemediumGemini 3.8 FlashlowvsQwen3.5-27BmediumGemini 3.8 FlashlowvsGPT-5.6 LunamediumGemini 3.8 FlashlowvsQwen3.5 Plus 2026-02-15mediumGemini 3.8 FlashlowvsLongCat 2.0medium

Category Breakdown

Category Score Consistency Tests Correct
Anti-AI Tricks 10.0 10.0
Coding 5.4 7.2
Combined 3.8 5.8
Data parsing and extraction 10.0 10.0
Domain specific 5.8 4.4
General Intelligence 10.0 10.0
Instructions following 10.0 10.0
Puzzle Solving 10.0 10.0
Tool Calling 10.0 10.0
Trivia 10.0 10.0

Compared models