Gemini 3.7 Flash

Gemini 3.7 Flash ranks first on 2 of the 209 benchmarks with five or more models scored.

Google DeepMind Released August 13th, 2026 Family Gemini 3.7 The lab's model card Compare with others

223

Benchmarks with a score

Ranked first, out of five or more models

In the top three

Google DeepMind Aug 2026

Gemini 3.7 Flash

Closed weights Mid-size

158.4 #25 of 256

Epoch Capabilities Index

Self-Reported Index 159.1 #38 of 404

DeepSWE 65.5% #12 /28
AutomationBench 30.4% #9 /42
Terminal-Bench 4.0 11.2% #33 /40

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Knowledge and language

Benchmark	Score	Rank	Source
OpenAI MRCR v2	89%	#1 of 47	Official leaderboard
LMArena (Elo)	1,489	#14 of 335	Official leaderboard
StudyArena Text	1,339	#4 of 85	Official leaderboard
Hemingway-bench	1,105	#2 of 42	Official leaderboard
MRCR v2	97%	#2 of 30	Lab model card
Judgemark v4	83.4	#4 of 40	Official leaderboard
Extended NYT Connections	94%	#11 of 103	Official leaderboard
SimpleQA Verified	73.2%	#6 of 56	Official leaderboard
FACTS Parametric	78.7%	#5 of 45	Official leaderboard
Global MMLU Lite	94.9%	#6 of 43	Official leaderboard
DesignArena ASCII Art	1,268	#14 of 77	Official leaderboard
Antidote	1,075	#7 of 36	Official leaderboard
Creative Writing v3	76.1	#29 of 120	Official leaderboard
Continual Learning Exam	33.7%	#3 of 11	Official leaderboard
Document Arena	1,468	#17 of 39	Official leaderboard
HLE Extended	31.6%	#16 of 31	Official leaderboard
BullshitBench V2	35%	#72 of 129	Official leaderboard
FACTS Search V2	80.9%	#12 of 21	Official leaderboard
BullshitBench	23.6%	#65 of 107	Official leaderboard
Lech Mazur Creative Writing	-0.76	#33 of 50	Official leaderboard
AA-LCR Extended	48.9%	#36 of 49	Official leaderboard
HLE-Verified	53.6%	#3 of 4	Lab model card

Math and science

Benchmark	Score	Rank	Source
GPQA Diamond	94.8%	#7 of 229	Epoch AI run
DTBench	96.8%	#12 of 180	Official leaderboard
ForecastBench	0.12	#7 of 82	Official leaderboard
ARC-AGI-2	84.6%	#16 of 98	Official leaderboard
OTIS Mock AIME 2024-2025	97.2%	#40 of 212	Epoch AI run
ARC-AGI	95.5%	#18 of 95	Official leaderboard
Human Behavior Bench	79.7	#3 of 11	Official leaderboard
FrontierMath Tiers 1-3 v2	71.6%	#27 of 86	Official leaderboard
BioMysteryBench	87.1%	#4 of 10	Lab model card
FrontierMath Tier 4 v2	36.6%	#28 of 68	Official leaderboard
ProofBench v1.1	58%	#26 of 52	Official leaderboard
LatchBio Capabilities v1.0	40.8%	#17 of 27	Official leaderboard
xbench-ScienceQA	70.2%	#20 of 29	Official leaderboard
Verifying the Verifiers (Project APE CRED benchmark)	74%	#16 of 23	Official leaderboard
scBench (LatchBio SingleCellBench)	50.4%	#20 of 26	Official leaderboard
FrontierChallenge	10.3%	#9 of 11	Official leaderboard
Terminal-Bench-Science	5.71%	#15 of 17	Official leaderboard
LABBench2	82.1%	#3 of 3	Lab model card

Coding

Benchmark	Score	Rank	Source
DesignArena Data Visualization	1,310	#13 of 118	Official leaderboard
DesignArena Website	1,296	#20 of 151	Official leaderboard
DesignArena Game Dev	1,313	#21 of 143	Official leaderboard
DesignArena SVG	1,280	#17 of 103	Official leaderboard
DesignArena Code	1,299	#24 of 143	Official leaderboard
DesignArena Android (Kotlin)	1,233	#8 of 46	Official leaderboard
DesignArena 3D Design	1,291	#28 of 136	Official leaderboard
WebDev Arena	1,587	#22 of 105	arena.ai, via Epoch AI
DesignArena React Native	1,228	#12 of 56	Official leaderboard
DesignArena UI Component	1,288	#30 of 139	Official leaderboard
ALE-Bench	1,876	#28 of 126	Official leaderboard
DesignArena Video to Website	1,213	#5 of 19	Official leaderboard
Image-to-WebDev Arena	1,600	#15 of 55	Official leaderboard
Code Arena	1,588	#35 of 128	Official leaderboard
DesignArena Image to HTML	1,227	#15 of 51	Official leaderboard
DesignArena Agentic Game Dev	1,182	#8 of 27	Official leaderboard
IOI v2 (Vals)	67.8%	#15 of 45	Official leaderboard
DesignArena Godot Game Dev	1,239	#8 of 24	Official leaderboard
DesignArena Agentic Overall	1,216	#25 of 74	Official leaderboard
DesignArena Frontend Web App	1,210	#22 of 53	Official leaderboard
Vibe Code Bench	70.4%	#50 of 120	Official leaderboard
DesignArena Full Stack	1,188	#21 of 49	Official leaderboard
DeepSWE	65.5%	#12 of 28	Official leaderboard
APEX-SWE	42%	#30 of 65	Official leaderboard
ProgramBench almost solved	5.50%	#12 of 25	Official leaderboard
ProgramBench tests passed	61.2%	#12 of 25	Official leaderboard
Code Migration	34.8%	#38 of 78	Official leaderboard
Vision2Web	76.5%	#5 of 10	Official leaderboard
MirrorCode	41.1%	#23 of 39	Epoch AI run
FrontierSWE	20.3%	#18 of 25	Official leaderboard
SWE-bench Verified Extended	30%	#34 of 44	Official leaderboard
Vals RSI Index	0.18	#22 of 27	Official leaderboard
FrontierCS 1.1	48.0	#5 of 6	Official leaderboard
FrontierCode 1.1	43.6%	#11 of 13	Lab model card
Senior SWE-Bench	14.7%	#21 of 23	Official leaderboard
SWE-Marathon v1.1	22.5%	#19 of 20	Official leaderboard
ProgramBench	0%	#25 of 25	Official leaderboard
RSI-Exam	30.9%	#18 of 18	Official leaderboard
Android Bench 2.0	7.30%	#17 of 17	Official leaderboard
Google RE-Bench (GRB)	27%	#1 of 1	Lab model card

Agents and tools

Benchmark	Score	Rank	Source
Chess Puzzles	66%	#7 of 93	Epoch AI run
Game Arena Coin Game	484	#3 of 39	Official leaderboard
GameHorizon	76.7	#3 of 37	Official leaderboard
Game Arena Chess	1,397	#4 of 42	Official leaderboard
LLM Chess	1,107	#19 of 123	Official leaderboard
BabaIsBench	100%	#5 of 32	Official leaderboard
RuneBench	5.96	#12 of 76	Official leaderboard
Terminal-Bench 2.1 Extended	44.4%	#7 of 44	Official leaderboard
Game Arena Reversi	818	#7 of 39	Official leaderboard
AutomationBench	30.4%	#9 of 42	Official leaderboard
Game Arena Four in a Row	552	#11 of 43	Official leaderboard
Game Arena Poker	6.46	#8 of 31	Official leaderboard
Terminal-Bench 2.1	85.8%	#15 of 56	Lab model card
Mystery Game Puzzles	37%	#23 of 84	Epoch AI run
Game Arena Word Association	329	#11 of 40	Official leaderboard
Kaggle Game Arena	371	#13 of 42	Official leaderboard
Game Arena Checkers	1,075	#13 of 41	Official leaderboard
Game Arena Bargaining	214	#13 of 39	Official leaderboard
Game Arena Word Art	1,072	#13 of 39	Official leaderboard
Game Arena Clobber	390	#14 of 39	Official leaderboard
Game Arena Dark Hex	292	#15 of 41	Official leaderboard
Game Arena Nine Men's Morris	711	#16 of 39	Official leaderboard
CLASHarness	21.5%	#6 of 14	Official leaderboard
Game Arena Go	1,006	#17 of 39	Official leaderboard
BrowseComp Extended	25.5%	#21 of 42	Official leaderboard
Game Arena Dots and Boxes	901	#20 of 40	Official leaderboard
TraceTrial-Bench	74.5%	#5 of 9	Official leaderboard
Game Arena Lines of Action	351	#22 of 39	Official leaderboard
Game Arena Ultimate Tic-Tac-Toe	267	#22 of 39	Official leaderboard
Agent Arena	-2.20%	#36 of 56	Official leaderboard
ApprenticeBench	16%	#12 of 18	Official leaderboard
Learn2Play Bench	58.3	#8 of 11	Official leaderboard
OSWorld 2.0	47.9%	#15 of 19	Lab model card
Terminal-Bench 4.1	12.4%	#35 of 43	Official leaderboard
Terminal-Bench 4.0	11.2%	#33 of 40	Official leaderboard
RecreationBench	21.1%	#9 of 10	Official leaderboard
Terminal-Bench 3.0	14.9%	#6 of 6	Lab model card

Professional work

Benchmark	Score	Rank	Source
LegalBench	87.3%	#6 of 138	Official leaderboard
Vals Finance Agent v2	59%	#7 of 80	Official leaderboard
MedCode	53.4%	#9 of 96	Official leaderboard
ProfBench	57.5%	#8 of 72	Official leaderboard
TRLawBench	9.61	#4 of 34	Official leaderboard
APEX-Agents 1.1	79.4%	#8 of 65	Official leaderboard
APEX-Agents	67.8%	#9 of 61	Collected by Epoch AI, source not named
Excel Modeling Benchmark	71.3%	#13 of 76	Official leaderboard
MedXpertQA MM Extended	62.7%	#9 of 41	Official leaderboard
TaxEval v2	74.7%	#32 of 129	Official leaderboard
Vals Index v2	59.3%	#18 of 65	Official leaderboard
APEX-1	67%	#21 of 71	Official leaderboard
Harvey LAB (Legal Agent Benchmark)	8.75%	#25 of 80	Official leaderboard
EnterpriseBench: CoreCraft	55.9%	#16 of 51	Official leaderboard
MortgageTax	66.7%	#30 of 87	Official leaderboard
MedScribe	83.9%	#38 of 97	Official leaderboard
Tuesday Work Index	47.9	#23 of 44	Official leaderboard
PolicyBench	82.7%	#27 of 50	Official leaderboard
Legal Research Bench	34.6%	#51 of 93	Official leaderboard
Vals Index v2.1	51.3%	#27 of 49	Official leaderboard
DesignArena Spreadsheet	1,121	#20 of 35	Official leaderboard
EngiWorld	26.1%	#4 of 7	Official leaderboard
Remote Labor Index	5.00%	#11 of 19	Official leaderboard
GDPval Extended	0.90%	#25 of 43	Official leaderboard
APEX-Accounting	57.2%	#29 of 49	Official leaderboard
GDPval-AA v2	1,525	#16 of 27	Lab model card
Tax Agent Bench	22.8%	#50 of 84	Official leaderboard
DAYJOB: Finance	5.30%	#24 of 39	Official leaderboard
DAYJOB: Healthcare	2.40%	#23 of 37	Official leaderboard
DesignArena PDF	1,163	#24 of 37	Official leaderboard
TaxCalcBench v2	26%	#22 of 32	Official leaderboard
$OneMillion-Bench	22.5%	#9 of 13	Official leaderboard
LifeSci Diamond	1.80%	#13 of 18	Official leaderboard
ATLAS-Finance	3.33%	#8 of 11	Official leaderboard
ATLAS-Finance v1.1.0	3.33%	#8 of 11	Official leaderboard
Agents' Last Exam	26.3%	#19 of 26	Lab model card
DesignArena Python-PPTX Slides	1,065	#32 of 36	Official leaderboard
Harvey LAB-AA v1.0	90.7%	#2 of 2	Lab model card

Vision, audio and robotics

Benchmark	Score	Rank	Source
1H-VideoQA	89.1%	#1 of 20	Official leaderboard
Vision Arena	1,294	#7 of 136	Official leaderboard
LVBench	85.4%	#3 of 31	Lab model card
CharXiv	88.7%	#13 of 75	Lab model card
Chartography	40.4%	#8 of 43	Official leaderboard
MMAU-Pro	70.4%	#2 of 10	Official leaderboard
ParseBench	70.4%	#9 of 44	Official leaderboard
ZeroBench	30%	#17 of 76	Official leaderboard
ClockBench	53.3%	#9 of 37	Official leaderboard
Humanity's Sixth Sense	39.8%	#10 of 34	Official leaderboard
MineBench	1,748	#23 of 75	Official leaderboard
VoxelBench	1,876	#18 of 56	Official leaderboard
DesignArena Code to Video	1,198	#10 of 29	Official leaderboard
GDP.pdf	34%	#6 of 17	Lab model card
MMMU-Pro Extended	36.4%	#14 of 38	Official leaderboard
vCX Benchmark	88.8%	#16 of 42	Official leaderboard
GlobeBench	21.8%	#9 of 19	Official leaderboard
CharXiv Extended	79.1%	#21 of 38	Official leaderboard
Code2Video Benchmark	1,477	#12 of 19	Official leaderboard
Furniture Assembly Benchmark	26.7%	#30 of 35	Official leaderboard
LADDER	42.8%	#2 of 2	Official leaderboard

Safety and security

Benchmark	Score	Rank	Source
nocot-bench (NCRI)	126.1	#6 of 185	Official leaderboard
BioSecBench-Refusal	54.8%	#3 of 22	Official leaderboard
World-Class Bio	58%	#11 of 74	Official leaderboard
Human Pathogen Capabilities Test	64.7%	#14 of 86	Official leaderboard
Molecular Biology Capabilities Test	55%	#21 of 90	Official leaderboard
VCT-v2	48.6%	#7 of 25	Official leaderboard
BioTIER Refuse	78.2%	#20 of 69	Official leaderboard
Human Pathogen Capabilities Test v2 (HPCT v2)	63.6%	#21 of 71	Official leaderboard
TxBench-Antibody-Discovery	50.7%	#10 of 30	Official leaderboard
GDM Stealth	75%	#2 of 6	Lab model card
GDM Situational Awareness	54.5%	#3 of 8	Lab model card
TxBench-Oligo-Discovery	37.8%	#13 of 30	Official leaderboard
SpatialBench-Long	29.8%	#12 of 27	Official leaderboard
SRE-Bench	4.58%	#16 of 35	Official leaderboard
scBench-Long	18.8%	#13 of 27	Official leaderboard
CBRN Conversational Evals	1.80%	#1 of 2	Lab model card
CBRN Science Eval	3.50%	#1 of 2	Lab model card
PCT (Polarized Contrastive Pairs)	45.1%	#24 of 46	Official leaderboard
TxBench-Preclinical-Pharmacology	55.3%	#16 of 30	Official leaderboard
EpiBench	22%	#15 of 27	Official leaderboard
BioTIER Permit	99.3%	#40 of 69	Official leaderboard
MetagenomicsBench	31.7%	#16 of 26	Official leaderboard
BioSecBench-Function	33.7%	#17 of 27	Official leaderboard
VariantBench	31.4%	#17 of 27	Official leaderboard
DistressBench	69%	#27 of 42	Official leaderboard
BioSecBench-Surveillance	41.5%	#18 of 27	Official leaderboard
OncologyBench	14.7%	#21 of 30	Official leaderboard
SecIT Bench	65.1%	#23 of 26	Official leaderboard
CyberBench v1.1 (Vals)	43.8%	#39 of 44	Official leaderboard
Chem / Bio Dynamic Jailbreak Eval	39%	#2 of 2	Lab model card
Cyber Adaptive Jailbreak Eval	24%	#2 of 2	Lab model card
Cyber Conversational Evals	11.4%	#2 of 2	Lab model card
Cyber Static Jailbreak Eval	50%	#2 of 2	Lab model card
Gray Swan Bio Eval	15%	#2 of 2	Lab model card
Gray Swan Cyber Eval	88%	#2 of 2	Lab model card
Harmful manipulation propensity evaluation	21.3%	#1 of 1	Lab model card

Other

Benchmark	Score	Rank	Source
CAIS Vision Capabilities Index	68.7	#8 of 53	Official leaderboard
ACCoRD	77.0	#24 of 152	Official leaderboard
CAIS Text Capabilities Index	49.3	#11 of 68	Official leaderboard
Conceptual Reasoning Index	64.6	#25 of 151	Official leaderboard
LMCA	50.4%	#28 of 147	Official leaderboard
Xbench	41.5	#4 of 17	Official leaderboard
SAGE (Vals)	49.2%	#23 of 81	Official leaderboard
CompassRank	80.5	#5 of 14	Official leaderboard
StudentBench	13.6	#5 of 11	Official leaderboard
CAIS Risk Index	51.1	#34 of 67	Official leaderboard
Artificial Analysis Intelligence Index v4.1	56.0	#5 of 8	Lab model card

Its lab's numbers put it at 159.1, in line with its ECI

The Self-Reported Index fits one score to the 11 capability results Google DeepMind reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 155.6 to 162.5. How we compute it.

Its lab's numbers next to 3 leaderboards

Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.

Benchmark	Lab says	Board says	Gap	What differs
DeepSWE	65.3% DeepSWE v1.1	65.5% Leaderboard, medium, mini-swe-agent	−0.2 pts
AutomationBench	30.4% AutomationBench (Private set)	30.4% Leaderboard, high	Same
Code Arena	1,588 Code Arena (Web development, Elo)	1,588 Leaderboard, high	Same	Google quotes an August Arena rating as an integer, while the board supplies a later rating with decimal precision.

Google DeepMind's models over time

Google DeepMind's models in color, against every other model on the Epoch Capabilities Index, by release date.

Source: Epoch Capabilities Index, Epoch AI (CC BY 4.0)

Its launch post

The benchmarks the lab chose to headline, with the scores it gave.

Gemini 3.7 Flash model card Aug 13, 2026

Artificial Analysis Intelligence Index 56
FrontierCode 1.1 Main (Score) 43.6
DeepSWE v1.1 65.3
Code Arena (Web development, Elo) 1588
Terminal-bench 2.1 85.8
Terminal-bench 3.0 14.9
AutomationBench (Private set) 30.4
GDPVal-AA v2 (Elo) 1525
Harvey LAB-AA 90.7
GDP.pdf 34
CharXiv Reasoning (With tools; No tools = 84.5%) 88.7
LVBench 85.4
GDM-MRCR v2 (8-needle), 128k (average) 97
OSWorld-2.0 47.9
Agent's Last Exam (Pass rate) 26.3
HLE-Verified 53.6
Best of BioMysteryBench, Human solvable / BioMysteryBench, Human difficult 87.1
LABBench2 82.1

Its system cards

Every result in the longer documents the lab published with it. 16 of them aren't in the launch post.

Frontier Safety Framework Report - Gemini 3.7 Flash Aug 13, 2026

Robustness