AI BENCHY
Advertise here
#41

Gemini 3 Flash Preview

Google Release: 2025-12-17 Tested on: 2026-05-22 00:23 google/gemini-3-flash-preview::none
~500B total (~20B active)MoEClosedEstimated
(medium) (low) (none)

Summary

Gemini 3 Flash Preview scores 7.7 on AI BENCHY and ranks #41. It has 10.0 reliability, a 70.0% pass rate, $0.023 total cost, and 1.70s average response time.

What makes Gemini 3 Flash Preview unique: It stands out most in Domain specific, where it ranks #1, while Coding is its weakest area at #18. Its total benchmark cost is unusually low for its score range. It is notably fast compared with similar models.

Model facts

Researched on 2026-08-12

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

Best estimate from public evidence; the vendor did not disclose every value. Sparse MoE is documented for the family; counts are estimated.

Parameter source Availability source

Score

7.7

Consistency

9.2

Reliability

10.0

Total Cost (Current Price)

$0.023

Total Output Tokens

1,879

Total Input Tokens

0

Input Price

$0.500 / 1M

Output Price

$3.000 / 1M

Tests Correct

Wrong Tests: 7

Attempt pass rate: 70.0%

Flaky tests

2

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

Response Time (avg)

1.70s

Response Time (max): 3.56s

Response Time (total): 22.06s

Wrong answer: 7

Hamster playing table tennis

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

#41 Gemini 3 Flash Preview

none
Cost
$0.003
Time
5.6s
Tokens
981 tok

Run history

Tested on Score Reliability Tests Correct Total Cost Compare
2026-08-14 02:39 Re-test 6.8 10.0 $0.085 Compare
2026-07-16 22:49 New test added 6.8 10.0 $0.085 Compare
2026-06-04 13:56 New test added 7.4 10.0 $0.025 Compare
2026-05-22 00:23 Suite changed 7.7 10.0 $0.023 Current run
2026-04-11 01:44 First recorded run 8.1 N/A $0.021 Compare

This run used a different benchmark suite. Keep suite changes in mind when reading historical movement.

Run comparison

RunBenchmark coverageScoreConsistencyReliabilityTests CorrectFlaky testsTotal Output TokensTotal Input TokensTotal CostResponse Time (avg)
2026-05-22 00:23 · Suite changed60/60 attempts7.79.210.013/2021,8790$0.0231.70s
2026-06-04 13:56 · New test added63/63 attempts7.49.210.013/2121,88537,011$0.0251.65s
Difference—+0.20.00.000-6-37011-$0.002+49ms

Benchmark coverage differs: 60/60 attempts (Target: 3 repeats per test) versus 63/63 attempts (Target: 3 repeats per test). Totals and repeat-sensitive metrics are not directly comparable.

These two runs used different benchmark suites, so the deltas reflect both model changes and suite changes.

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 Flash PreviewnonevsMiMo-V2.5-PromediumGemini 3 Flash PreviewnonevsGemma 4 26B A4BmediumFree AvailableGemini 3 Flash PreviewnonevsDeepSeek V4 PrononeGemini 3 Flash PreviewnonevsGLM 5.2noneGemini 3 Flash PreviewnonevsSolar Pro 4mediumGemini 3 Flash PreviewnonevsKAT-Coder-Pro V2.5noneGemini 3 Flash PreviewnonevsSolar Pro 4lowGemini 3 Flash PreviewnonevsLongCat 2.0lowGemini 3 Flash PreviewnonevsSeed-2.0-CodemediumGemini 3 Flash PreviewnonevsGemini 3.5 Flashminimal

Category Breakdown

Category Score Consistency Tests Correct
Anti-AI Tricks 8.3 10.0
Coding 6.8 10.0
Combined 4.7 1.6
Data parsing and extraction 10.0 10.0
Domain specific 7.7 10.0
General Intelligence 10.0 10.0
Instructions following 6.4 5.8
Puzzle Solving 7.7 10.0
Tool Calling 10.0 10.0
Trivia 3.0 10.0

Compared models