Google DeepMindmodel systemcard
Gemini 3.1 Pro
#11of 27 · general
preview builds collapsed
1Mcontext
Proprietarynon-commercial
2026-02-19released
2025-01-31knowledge cutoff
Standing
dot = this model · shaded = the field's middle 60% · track = full fieldGeneral
54.6#11/27
Math
70.0#6/27
Coding
34.4#17/27
Health
52.5#12/26
Cyber
0.0#14/14
Legal
43.3#17/26
Instruct
93.1#1/27
Agents
36.0#14/27
Professional
53.8#11/27
Multimodal
42.2#9/19
Hallucination
55.0#4/27
Speed
39.0#14/22
Value
48.4#13/27
Measured
≈ imputed · amber dot = lab-self-reported among the sourcesGPQA Diamond94.4%#2
HLE47.0%#8
MMLU-Pro91.0%#3
AIME 202598.1%#1
LiveBench78.4%#6
SWE-bench V78.8%#12
LiveCodeBench88.5%#5
SciCode58.9%#2
CritPt17.7%#9
Terminal-Bench70.8%#12
APEX Agents33.5%#11
GDPval23.2%#20
Arena1486#4
AA Index47.7#14
Vals Index53.8#18
Epoch Capabilities Index154.8#10
Operations
$2 / $12$ per 1M tokens, in / out
$4.5blended
126tokens / sec
$0.29$ per solved task
1486arena elo · 86,683 votes