Google DeepMindmodel systemcard
Gemini 3.5 Flash
#15of 27 · general
high/medium effort collapsed
1Mcontext
Proprietarynon-commercial
2026-05-19released
2026-01-31knowledge cutoff
Standing
dot = this model · shaded = the field's middle 60% · track = full fieldGeneral
47.4#15/27
Math
50.6#11/27
Coding
33.9#18/27
Health
30.9#19/26
Cyber
46.2#8/14
Legal
6.2#26/26
Instruct
71.1#3/27
Agents
30.1#18/27
Professional
51.6#12/27
Multimodal
62.2#5/19
Hallucination
49.5#6/27
Speed
92.4#2/22
Value
44.8#15/27
Measured
≈ imputed · amber dot = lab-self-reported among the sourcesGPQA Diamond92.7%#9
HLE40.8%#15
MMLU-Pro89.5%#4
LiveBench74.8%#11
SWE-bench V79.1%#10
LiveCodeBench87.6%#7
SciCode53.1%#11
CritPt12.0%#16
Terminal-Bench75.2%#10
Arena1476#9
AA Index46.7#15
Vals Index62.8#12
Epoch Capabilities Index154.6#12
Operations
$1.5 / $9$ per 1M tokens, in / out
$3.38blended
174tokens / sec
$0.25$ per solved task
1476arena elo · 10,011 votes