modelbenchmark.io

Gemini 3 Flash Preview

Google · gemini-3-flash

Compare

New Gemini flash lane bringing frontier-style multimodal reasoning to cheaper runs

Specification

google listing

Context
1M
Max output
66K
Released
2025-12-17
Knowledge cutoff
2025-01
Retires
Open weights
no
Input
text, image, video, audio, pdf
Output
text

Price

US dollars per million tokens · google list price

Input
$0.50
Output
$3.00
Cache read
$0.05
Cache write
$0.0833

Batch and priority prices

US dollars per million tokens · per listing

PriceListing$/M
Batch inputvertex_ai$0.25
Batch outputvertex_ai$1.50
Priority inputvertex_ai$0.90
Priority outputvertex_ai$5.40

Batch input is 50% below the interactive input price at the same listing.

Rate limits

as listed

Requests per minute
2,000
Tokens per minute
800,000

Quality

10 benchmarks · 2 sources

Composite
84th
percentile of 334 scored models
Rank
53 / 334
±0.05 sd
Evidence
10 × 2
3 or more benchmarks, from 2 or more sources
Effort range
2.8
points between effort settings
coding82nd1/3 bench
reasoning91st3/4 bench
math70th5/7 bench
knowledge90th1/1 bench

By reasoning effort

every setting placed on the same scale as the leaderboard

SettingCompositeBenchmarks behind it
high92nd6
low74th1
minimal71st1

Every score

one row per source and configuration — nothing averaged away

BenchmarkScoreConfigurationSourceRun
Chess Puzzles40.0 ±4.9highEpoch AI2026-08-06
38.0 ±4.9defaultEpoch AI2025-12-17
Settings differ by 2.0 points on this benchmark.
SimpleQA Verified66.8 ±1.5highEpoch AI2026-08-27
OTIS Mock AIME 2024-202595.6 ±3.1highEpoch AI2026-08-06
92.8 ±3.5defaultEpoch AI2025-12-17
Settings differ by 2.8 points on this benchmark.
FrontierMath-v147.8 ±16.3default · 2 runsEpoch AI2025-12-17
GPQA diamond89.4 ±2.2highEpoch AI2026-08-06
83.2 ±1.9defaultEpoch AI2025-12-17
Settings differ by 6.2 points on this benchmark.
SWE-Bench verified75.8high · mini-SWE-agentSWE-bench
75.4 ±2.0defaultEpoch AI2026-02-18
Sources and settings disagree by 0.4 points. Both figures stand.
Mystery Game Puzzles26.0 ±4.4lowEpoch AI2026-08-05
25.0 ±4.4minimalEpoch AI2026-08-05
20.0 ±4.0highEpoch AI2026-08-05
Settings differ by 6.0 points on this benchmark.
FrontierMath-Tiers-1-3-v251.2 ±3.0defaultEpoch AI2026-06-11
FrontierMath-Tier-4-2025-07-012.1 ±2.9default · 2 runsEpoch AI2025-12-17
FrontierMath-Tier-4-v217.1 ±5.9defaultEpoch AI2026-06-11

The leading figure for a benchmark is the median across its configurations, so one heroic high-effort run cannot set the number.

Available from 35 hosts

HostIn $/MOut $/MCache rdCache wrContextOutputRetires
googlevendor$0.50$3.00$0.051M66K
qihang-ai$0.07$0.431M66K
kilo$0.25$1.50$0.025$0.04171M66K
poe$0.40$2.40$0.041M66K
302ai$0.50$3.001M66K
abacus$0.50$3.00$0.051M66K
aihubmix$0.50$3.00$0.051M66K
crossmodel$0.50$3.00$0.05$0.501M66K
databricks
$0.50
$0.625
$3.00
$3.75
$0.05
$0.0625
$0.6251M66K
edenai$0.50$3.00$0.05$0.08331M66K
frogbot$0.50$3.00$0.051M66K
gemini$0.50$3.00$0.051M66K
gmi · gmi$0.50$3.001M66K
google-vertex$0.50$3.00$0.051M66K
jiekou$0.50$3.001M66K
llmgateway$0.50$3.00$0.051M66K
llmgateway-providers · google-ai-studio$0.50$3.00$0.051M66K
llmgateway-providers · google-vertex$0.50$3.00$0.051M66K
merge-gateway$0.50$3.00$0.051M66K
nano-gpt$0.50$3.00$0.051M66K
neon$0.50$3.00$0.051M66K
ofox$0.50$3.00$0.05$1.001M66K
opencode$0.50$3.00$0.051M66K
openrouter$0.50$3.00$0.05$0.08331M
66K
66K
opper$0.50$3.00$0.051M66K
orcarouter$0.50$3.00$0.051M66K
perplexity$0.50$3.00$0.05
perplexity-agent$0.50$3.00$0.051M66K
pioneer$0.50$3.00$0.05$0.08331M65K
vercel$0.50$3.00$0.051M65K
vertex_ai$0.50$3.00$0.051M66K
vertex_ai-language-models$0.50$3.00$0.051M66K
zenmux$0.50$3.00$0.05$1.001M64K
venice$0.70$3.75$0.07256K66K
anyapi1M66K

More from Google

most-hosted first