Benchmark Almanac

Gemini 3.1 Pro

Gemini 3.1 Pro ranks first on 26 of the 402 benchmarks with five or more models scored.

Google DeepMind Released February 19th, 2026 Family Gemini 3.1 The lab's model card Compare with others
418
Benchmarks with a score
26
Ranked first, out of five or more models
76
In the top three
Google DeepMind Feb 2026

Gemini 3.1 Pro

Closed weights Flagship

156.2 #39 of 256
Epoch Capabilities Index
Self-Reported Index 157.5 #46 of 404
  • Humanity's Last Exam 46.4% #3 /47
  • SWE-bench Verified 75.6% #12 /32
  • SWE-Bench Pro 46.1% #5 /24

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

BenchmarkScoreRankSource
MMLU-Pro 91.2% #1 of 212 Official leaderboard
SEA-HELM 77.4 #1 of 71 Official leaderboard
SimpleQA Verified 77.5% #1 of 56 Official leaderboard
HUMAINE 35 #1 of 50 Official leaderboard
Global PIQA 93.9% #1 of 44 Official leaderboard
KINA 53.2% #1 of 37 Official leaderboard
MMMLU 92.6% #2 of 71 Lab model card
FACTS Search 85.6% #1 of 31 Official leaderboard
SkateBench v2 96.9% #1 of 21 Official leaderboard
Extended NYT Connections 97.4% #5 of 103 Official leaderboard
CulturalRiddles 79.4% #1 of 20 Official leaderboard
LMArena (Elo) 1,487 #17 of 335 Official leaderboard
Antidote 1,091 #2 of 36 Official leaderboard
Humanity's Last Exam 46.4% #3 of 47 Official leaderboard
MultiNRC 64.7% #3 of 38 Official leaderboard
Joke Arena 1,567 #2 of 25 Official leaderboard
FACTS Parametric 79% #4 of 45 Official leaderboard
STALE 55.2% #1 of 10 Official leaderboard
Structured Output Benchmark 86.9% #2 of 19 Official leaderboard
Last Translation Benchmark 39.3 #5 of 36 Official leaderboard
Hemingway-bench 1,073 #6 of 42 Official leaderboard
FACTS Parametric V2 79.4% #5 of 27 Official leaderboard
MultiChallenge 71.4% #8 of 42 Official leaderboard
Judgemark v4 78.7 #8 of 40 Official leaderboard
DesignArena ASCII Art 1,265 #16 of 77 Official leaderboard
EQ-Bench 3 1,527 #17 of 78 Official leaderboard
Visual Aesthetic Benchmark 35% #4 of 18 Official leaderboard
Corp Speak Arena 1,578 #5 of 22 Official leaderboard
MRCR v2 84.9% #7 of 30 Lab model card
CL-bench 20.8% #6 of 24 Official leaderboard
EvoArena Terminal-Bench-Evo 56.5% #2 of 8 Official leaderboard
Phare 70.6% #18 of 62 Official leaderboard
CL-bench Life 16.9% #5 of 16 Official leaderboard
FACTS Search V2 83.7% #8 of 21 Official leaderboard
FACTS Benchmark Suite 69.2% #9 of 23 Official leaderboard
FACTS Multimodal 50.4% #9 of 23 Official leaderboard
Creative Writing v3 71.2 #47 of 120 Official leaderboard
Longform Creative Writing (EQ-Bench) 68.2 #47 of 119 Official leaderboard
OpenAI MRCR v2 47.4% #19 of 47 Official leaderboard
ELI5 Arena 1,468 #8 of 18 Official leaderboard
StudyArena Text 1,234 #42 of 85 Official leaderboard
BullshitBench V2 37% #68 of 129 Official leaderboard
Document Arena 1,459 #23 of 39 Official leaderboard
Vectara Hallucination Leaderboard (2025 dataset) 10.4% #62 of 104 Official leaderboard
EvoArena PersonaMem-Evo 48.3% #5 of 8 Official leaderboard
BullshitBench 21.8% #67 of 107 Official leaderboard
HLE Extended 19.6% #21 of 31 Official leaderboard
EvoArena SWE-Chain-Evo 20.5% #6 of 8 Official leaderboard
Doubao Multi-Turn Bench 52% #3 of 4 Official leaderboard
EQ-Bench 4 1,132 #25 of 32 Official leaderboard
Lech Mazur Creative Writing -2.21 #41 of 50 Official leaderboard
FACTS Grounding v2 56.6% #21 of 25 Official leaderboard
AA-LCR Extended 35.4% #44 of 49 Official leaderboard
Continual Learning Bench 0.00 #5 of 5 Official leaderboard
xDailyBench 35.2% #4 of 4 Official leaderboard

Math and science

BenchmarkScoreRankSource
GPQA Diamond 94.4% #9 of 229 Epoch AI run
MathArena Apex Shortlist 92% #2 of 38 Official leaderboard
DTBench 97.1% #11 of 180 Official leaderboard
ARC-AGI 98% #9 of 95 Official leaderboard
MathArena (Apex, IMO, USAMO) 60.9% #5 of 47 Official leaderboard
SimpleBench 79.6% #13 of 106 Official leaderboard
AIME 2026 98.3% #5 of 31 Official leaderboard
ForecastBench 0.12 #14 of 82 Official leaderboard
FrontierMath 36.9% #11 of 55 Official leaderboard
IMO-ProofBench 49% #4 of 20 Official leaderboard
OTIS Mock AIME 2024-2025 95.6% #45 of 212 Epoch AI run
FrontierMath Tier 4 16.7% #12 of 55 Epoch AI run
IMProofBench 45.7% #2 of 9 Official leaderboard
ARC-AGI-2 77.1% #22 of 98 Official leaderboard
Echo General AI Prediction Leaderboard 1,006 #4 of 17 Official leaderboard
LiveMathematicianBench 37.6% #4 of 17 Official leaderboard
Project Euler (MathArena) 89% #1 of 4 Official leaderboard
HMMT 2026 94.7% #8 of 31 Official leaderboard
USAMO 2026 74.4% #3 of 9 Official leaderboard
AstaBench 49.6% #5 of 13 Official leaderboard
ProofBench 26% #25 of 59 Official leaderboard
FrontierMath Tiers 1-3 v2 59.6% #41 of 86 Official leaderboard
SpatialBench (LatchBio) 51.6% #7 of 14 Official leaderboard
HorizonMath 2.70% #4 of 8 Official leaderboard
FrontierMath Tier 4 v2 26.8% #38 of 68 Official leaderboard
Leipzig Benchmark 15% #12 of 19 Official leaderboard
Prophet Arena 85.5 #16 of 24 Official leaderboard
BTF-2 0.14 #2 of 3 Official leaderboard
ProofBench v1.1 26% #38 of 52 Official leaderboard
KellyBench -66.00% #4 of 5 Official leaderboard
MysteryMechanism 20.7% #26 of 31 Official leaderboard
BixBench3 24.8 #17 of 20 Official leaderboard
ResearchClawBench 13.3 #21 of 23 Official leaderboard
BioMysteryBench 60% #28 of 30 Official leaderboard
PhysClosureBench 2.85 #7 of 7 Official leaderboard
IMO-LeanProofBench 11.6% #1 of 1 Official leaderboard

Coding

BenchmarkScoreRankSource
SciCode 59% #2 of 49 Lab model card
LiveSQLBench 36.5% #3 of 36 Official leaderboard
AlgoTune 2.02× #2 of 18 Official leaderboard
METR Time Horizon 6.4 h #3 of 26 Claim in metr.org, via Epoch AI
METR Time Horizon 1.1 6.4 h #3 of 26 Official leaderboard
ALE-Bench 2,041 #19 of 126 Official leaderboard
WeirdML 72.1% #19 of 112 Official leaderboard
DesignArena SVG 1,276 #18 of 103 Official leaderboard
AutoLab 0.50 #2 of 10 Official leaderboard
SWE-Bench Pro 46.1% #5 of 24 Official leaderboard
LiveCodeBench Pro 2,887 #1 of 4 Lab model card
OpenHands Index 60.6% #9 of 33 Official leaderboard
FrontierCS 68.9 #4 of 14 Official leaderboard
ts-bench 60% #2 of 7 Official leaderboard
HarnessDev 55.6% #2 of 6 Official leaderboard
AndroidBench 74.3% #11 of 32 Official leaderboard
PostTrainBench 21.6% #8 of 23 Official leaderboard
RoadmapBench 20.9% #4 of 11 Official leaderboard
SWE-bench Verified 75.6% #12 of 32 Epoch AI run
DesignArena 3D Design 1,217 #53 of 136 Official leaderboard
FrontierCode 16.7% #5 of 12 Official leaderboard
SWE-Bench Pro Private 32.2% #5 of 12 Official leaderboard
GSO-Bench 22.6% #13 of 30 Official leaderboard
DAXBench v2 81.2% #21 of 47 Official leaderboard
DesignArena Website 1,227 #69 of 151 Official leaderboard
DesignArena Code 1,221 #67 of 143 Official leaderboard
DesignArena UI Component 1,216 #67 of 139 Official leaderboard
DesignArena Game Dev 1,192 #69 of 143 Official leaderboard
WebDev Arena 1,447 #54 of 105 arena.ai, via Epoch AI
DesignArena Data Visualization 1,207 #62 of 118 Official leaderboard
Code Arena 1,446 #69 of 128 Official leaderboard
IOI v2 (Vals) 51.8% #25 of 45 Official leaderboard
QwenVision2Code 1,632 #3 of 5 Official leaderboard
DesignArena Agentic Image to HTML 1,192 #7 of 11 Official leaderboard
ProgramBench 0% #16 of 25 Official leaderboard
Vibe Code Bench 32% #77 of 120 Official leaderboard
FrontierSWE v1 34.4% #11 of 17 Official leaderboard
SWE-Marathon 4.00% #8 of 12 Official leaderboard
Code Migration 17.3% #54 of 78 Official leaderboard
EvoCode-Bench 13.7% #9 of 13 Official leaderboard
SWE Atlas Refactoring 33.8% #12 of 17 Official leaderboard
DesignArena Godot Game Dev 1,145 #17 of 24 Official leaderboard
Image-to-WebDev Arena 1,480 #39 of 55 Official leaderboard
DesignArena Agentic Overall 1,080 #54 of 74 Official leaderboard
APEX-SWE 33.9% #48 of 65 Official leaderboard
DesignArena Frontend Web App 1,108 #40 of 53 Official leaderboard
DesignArena Full Stack 1,046 #37 of 49 Official leaderboard
Vision2Web S2-2026 28.8% #7 of 9 Official leaderboard
ReactBench 25.6% #22 of 28 Official leaderboard
SWE Atlas Codebase QnA Extended 2.70% #30 of 38 Official leaderboard
DesignArena Image to Frontend 1,117 #25 of 31 Official leaderboard
SWE Atlas Test Writing 29.8% #18 of 22 Official leaderboard
Data Agent Benchmark 60% #15 of 18 Official leaderboard
ProgramBench tests passed 36.4% #21 of 25 Official leaderboard
DesignArena Video to Website 1,146 #16 of 19 Official leaderboard
DesignArena React Native 1,088 #48 of 56 Official leaderboard
GBAEval 0.85% #18 of 21 Official leaderboard
SWE Atlas 13.5% #19 of 22 Official leaderboard
DesignArena Agentic Game Dev 1,062 #24 of 27 Official leaderboard
DesignArena Image to HTML 1,130 #46 of 51 Official leaderboard
PostTrainBench 1.1 22% #10 of 11 Official leaderboard
ScarfBench 37.6% #10 of 11 Official leaderboard
DesignArena Android (Kotlin) 1,035 #42 of 46 Official leaderboard
ProgramBench almost solved 0% #23 of 25 Official leaderboard
SWE-bench Verified Extended 21.3% #41 of 44 Official leaderboard
FrogsGame post-training task 0% #16 of 17 Official leaderboard
PostTrainBench 1.2 18.1% #20 of 21 Official leaderboard
MirrorCode 8.89% #38 of 39 Epoch AI run
DeepSWE 11.7% #28 of 28 Official leaderboard
Vibe Code Bench 1-100 6.69% #26 of 26 Official leaderboard
Senior SWE-Bench 2.11% #23 of 23 Official leaderboard
QwenBlenderBench 23% #6 of 6 Official leaderboard
MiMo Coding Bench 67.8% #5 of 5 Official leaderboard
RE-Bench 1.27 #1 of 1 Lab model card

Agents and tools

BenchmarkScoreRankSource
SnakeBench 37 #5 of 188 Official leaderboard
Terminal-Bench 2.0 80.2% #2 of 43 Official leaderboard
Game Arena Five in a Row 1,131 #1 of 20 Official leaderboard
PinchBench 80.7% #3 of 48 Official leaderboard
LLM Chess 1,511 #10 of 123 Official leaderboard
Clocktower Radio 1,706 #2 of 23 Official leaderboard
PlanBench-XL 77.1% #1 of 9 Official leaderboard
SkillsBench 60.8% #3 of 22 Official leaderboard
ADeptS-Bench Desktop 76% #1 of 7 Official leaderboard
ADeptS-Bench Mobile 73.8% #1 of 7 Official leaderboard
LisanBench 6445.0 #20 of 133 Official leaderboard
BALROG 57% #5 of 33 Official leaderboard
AutoResearchBench 7.93% #2 of 13 Official leaderboard
Game Arena Chess 1,314 #7 of 42 Official leaderboard
Pencil Puzzle Bench 33.3% #9 of 53 Official leaderboard
Toolathlon 48.8% #9 of 50 Official leaderboard
Game Arena Four in a Row 605 #8 of 43 Official leaderboard
Chess Puzzles 51% #18 of 93 Epoch AI run
Cooperate to Compete (C2C) 0.74 #1 of 5 Official leaderboard
OSWorld-Verified 78.3% #7 of 33 Official leaderboard
Game Arena Clobber 483 #9 of 39 Official leaderboard
BrowseComp 85.9% #19 of 81 Lab model card
GameHorizon 72.8 #9 of 37 Official leaderboard
Game Arena Werewolf -0 #2 of 8 Official leaderboard
MobileGym 58.8% #1 of 4 Official leaderboard
Game Arena Coin Game 420 #10 of 39 Official leaderboard
Kaggle Game Arena 384 #11 of 42 Official leaderboard
PinchBench 2.0 81% #15 of 57 Official leaderboard
PACT 1,557 #7 of 26 Official leaderboard
EnvSimBench 42% #2 of 7 Official leaderboard
Game Arena Word Association 316 #12 of 40 Official leaderboard
DeepSearchQA 60.2% #3 of 10 Official leaderboard
Game Arena Go 1,147 #12 of 39 Official leaderboard
Game Arena Nine Men's Morris 787 #12 of 39 Official leaderboard
Game Arena Word Art 1,103 #12 of 39 Official leaderboard
Mystery Game Puzzles 34% #27 of 84 Epoch AI run
Game Arena Dots and Boxes 966 #13 of 40 Official leaderboard
EnterpriseOps-Gym 38.7% #4 of 12 Official leaderboard
TUA-Bench 49.3% #4 of 12 Official leaderboard
Harbor-Index 13.4% #3 of 9 Official leaderboard
Game Arena Dark Hex 294 #14 of 41 Official leaderboard
Game Arena Reversi 710 #14 of 39 Official leaderboard
Claw-Eval 77.3% #9 of 24 Official leaderboard
LMArena Search Arena 1,208 #14 of 37 Official leaderboard
Game Arena Lines of Action 443 #15 of 39 Official leaderboard
Game Arena Checkers 982 #16 of 41 Official leaderboard
Bench on the Clocktower 52% #4 of 10 Official leaderboard
Toloka Arena 25.2% #2 of 5 Official leaderboard
DeepResearch Bench 47.8% #9 of 22 evals.futuresearch.ai, via Epoch AI
Deep Research Bench (FutureSearch) 0.48 #9 of 21 Official leaderboard
BabaIsBench 92% #14 of 32 Official leaderboard
Toolathlon-Verified 61.1% #11 of 25 Official leaderboard
OSWorld-MCP 44.2% #4 of 9 Official leaderboard
Terminal-Bench 2.1 Extended 35% #20 of 44 Official leaderboard
TERMS-Bench 0.64 #5 of 11 Official leaderboard
MCP Atlas 78.2% #17 of 37 Official leaderboard
Game Arena Bargaining 188 #18 of 39 Official leaderboard
DuelLab GameBench 2 65.0 #27 of 57 Official leaderboard
Game Arena Ultimate Tic-Tac-Toe 357 #19 of 39 Official leaderboard
GAIA 55.5% #32 of 64 Official leaderboard
UniClawBench 32.5% #5 of 10 Official leaderboard
CocoaBench 30.7% #3 of 6 Official leaderboard
Claw-Eval Pass^3 57.8% #13 of 24 Official leaderboard
DiG-bench 16.8% #4 of 7 Official leaderboard
Gaia2-CLI 52% #4 of 7 Official leaderboard
Claw-Eval-Live 53.3% #7 of 12 Official leaderboard
RuneBench 4.48 #45 of 76 Official leaderboard
ScienceBoard 11.8% #9 of 15 Official leaderboard
AdaPlanBench 34.5% #6 of 10 Official leaderboard
MobileWorld 38.8% #11 of 18 Official leaderboard
τ³-bench Banking 26% #16 of 25 Official leaderboard
TextQuests 18.8 #18 of 28 Official leaderboard
AutomationBench 8.68% #29 of 42 Official leaderboard
Agent Arena -4.50% #40 of 56 Official leaderboard
EBR-bench 14.3% #23 of 31 Epoch AI run
WildClawBench 40.8% #23 of 31 Official leaderboard
Game Arena Poker -16.06 #24 of 31 Official leaderboard
ARC-AGI-3 0.42% #21 of 27 Official leaderboard
ClawProBench 54% #39 of 49 Official leaderboard
Long-Horizon-Terminal-Bench (LHTB) 27.9% #20 of 25 Official leaderboard
OSWorld-Science 17.2% #9 of 11 Official leaderboard
BrowseComp Extended 10.7% #35 of 42 Official leaderboard
MM-ClawBench 61.8% #5 of 6 Official leaderboard
DeepTerminalBench 16% #8 of 9 Official leaderboard
WeaveBench 1.80% #8 of 9 Official leaderboard
Terminal-Bench 2.1 65.8% #12 of 13 Official leaderboard
CLASHarness 12% #13 of 14 Official leaderboard
Long-Horizon-Terminal-Bench 2.0 41.3% #15 of 15 Official leaderboard
SeedClawBench 57.1% #5 of 5 Official leaderboard
CreativeWork 27.4% #4 of 4 Official leaderboard

Professional work

BenchmarkScoreRankSource
MedCode 59.1% #2 of 96 Official leaderboard
TRLawBench 9.85 #1 of 34 Official leaderboard
LegalBench 87.4% #5 of 138 Official leaderboard
MortgageTax 69.4% #6 of 87 Official leaderboard
Vals Finance Agent 59.7% #7 of 49 Official leaderboard
OccuBench 73.3% #2 of 14 Official leaderboard
3DCodeArena 1,116 #3 of 14 Official leaderboard
CaseLaw v2 64.8% #12 of 53 Official leaderboard
TaxCalcBench 49% #4 of 17 Official leaderboard
ProfBench 55.2% #17 of 72 Official leaderboard
Vals Index v1 58.1% #9 of 34 Official leaderboard
MedXpertQA MM Extended 60.8% #11 of 41 Official leaderboard
CorpFin v2 64.5% #37 of 119 Official leaderboard
RadLE 2.0 660.0 #4 of 12 Official leaderboard
prinzbench 50.0 #9 of 26 Official leaderboard
SpreadsheetBench 2 23.7% #3 of 8 Official leaderboard
TaxEval v2 72.9% #56 of 129 Official leaderboard
MMEnterprise-Bench 62.5% #4 of 9 Official leaderboard
MBABench 16.1 #3 of 6 Official leaderboard
FORTE 70.3% #11 of 21 Official leaderboard
CADGenBench 31.1% #6 of 11 Official leaderboard
APEX-Agents 33.5% #7 of 12 Lab model card
PolicyBench 81.2% #30 of 50 Official leaderboard
BankerToolBench v1.1 13% #6 of 10 Official leaderboard
PRBench 44% #29 of 48 Official leaderboard
Physical-Engineering's Last Exam 0.15 #11 of 18 Official leaderboard
APEX-1 61.8% #45 of 71 Official leaderboard
Vending-Bench 2 $3,774 #44 of 67 Official leaderboard
Vals Index v2 41.9% #43 of 65 Official leaderboard
EEBench 46.3% #21 of 30 Official leaderboard
Agents' Last Exam 16.4% #22 of 31 Official leaderboard
Excel Modeling Benchmark 52.6% #54 of 76 Official leaderboard
DesignArena HTML Slides 1,048 #20 of 28 Official leaderboard
MedScribe 76.1% #70 of 97 Official leaderboard
Vals Finance Agent v2 43% #59 of 80 Official leaderboard
Legal Research Bench 20.7% #69 of 93 Official leaderboard
OfficeQA Pro 39.1% #9 of 12 Official leaderboard
PatientAgentBench 3.71 #6 of 8 Official leaderboard
EnterpriseBench: CoreCraft 27.2% #39 of 51 Official leaderboard
GDPval Extended 0.20% #34 of 43 Official leaderboard
Tax Agent Bench 9.29% #70 of 84 Official leaderboard
LifeSci Diamond 0% #15 of 18 Official leaderboard
HealthBench Professional 43.8% #5 of 6 Official leaderboard
VWE-Bench 42.7% #5 of 6 Official leaderboard
Harvey LAB (Legal Agent Benchmark) 0% #67 of 80 Official leaderboard
DAYJOB: Healthcare 0.40% #31 of 37 Official leaderboard
E-Commerce Bench 1.30 #16 of 19 Official leaderboard
APEX-Agents 1.1 52.2% #55 of 65 Official leaderboard
DAYJOB: Finance 0.30% #33 of 39 Official leaderboard
GDP.xlsx 10.3% #17 of 20 Official leaderboard
OfficeQA Pro V2 14.4% #6 of 7 Official leaderboard
Public Benefits Bench 53.8% #43 of 50 Official leaderboard
Workspace-Bench-Lite 37.5% #8 of 9 Official leaderboard
DesignArena PDF 1,104 #33 of 37 Official leaderboard
TaxCalcBench v2 2.00% #29 of 32 Official leaderboard
Tuesday Work Index 29.5 #40 of 44 Official leaderboard
CHI-Bench 7.10% #21 of 23 Official leaderboard
Vals Index v2.1 33.4% #45 of 49 Official leaderboard
PhysicianBench 6.00% #12 of 13 Official leaderboard
DesignArena Spreadsheet 1,007 #33 of 35 Official leaderboard
DesignArena Python-PPTX Slides 1,027 #34 of 36 Official leaderboard
YC-Bench $66,104 #56 of 59 Official leaderboard
FrontierFinance 30.7% #17 of 17 Official leaderboard
GDPval-AA 1,317 #16 of 16 Lab model card
JobBench 15.9% #16 of 16 Official leaderboard
ATLAS-Finance 0.67% #11 of 11 Official leaderboard
ATLAS-Finance v1.1.0 0.67% #11 of 11 Official leaderboard
Management Consulting Tasks (OpenAI) 13.2% #7 of 7 Official leaderboard
Agent Startup Bench 45.7% #5 of 5 Official leaderboard

Vision, audio and robotics

BenchmarkScoreRankSource
MATH-Vision (MathVision) 95.7% #2 of 105 Official leaderboard
ChartMuseum 80.7% #1 of 21 Official leaderboard
CiteVQA 76% #1 of 16 Official leaderboard
OVO-S-Bench 59.2% #1 of 15 Official leaderboard
VisFactor 54% #2 of 26 Official leaderboard
MMAU-Pro 72.2% #1 of 10 Official leaderboard
MeasureBench Real-world 36.1% #2 of 19 Official leaderboard
SpatialWorld Digital 39% #1 of 9 Official leaderboard
WorldBench 64% #1 of 8 Official leaderboard
VoxelBench Image 1,535 #2 of 13 Official leaderboard
MeasureBench Synthetic 33.2% #3 of 19 Official leaderboard
Vision Arena 1,280 #22 of 136 Official leaderboard
WorldSense (omnimodal) 65.5% #2 of 12 Official leaderboard
OmniCap-IF 80.7% #1 of 6 Official leaderboard
MMMU-Pro 80.5% #16 of 90 Lab model card
GeoBench 84% #9 of 48 Official leaderboard
Daily-Omni 82.8% #3 of 14 Official leaderboard
VoxMem 35.9% #2 of 9 Official leaderboard
Math Kangaroo 2025 (MathArena) 89.4% #5 of 22 Official leaderboard
ZeroBench 29% #18 of 76 Official leaderboard
PerceptionBench (Moonshot) 56.2% #4 of 16 Official leaderboard
Image2FloorPlan 55.1 #1 of 4 Official leaderboard
Omni-InteractScope Proactive 46.4% #1 of 4 Official leaderboard
QwenVisualOffice 39.6% #2 of 6 Official leaderboard
MineBench 1,661 #27 of 75 Official leaderboard
ParseBench 67.3% #17 of 44 Official leaderboard
1H-VideoQA 78.2% #8 of 20 Official leaderboard
CC-OCR-Bench-V2 70.2% #6 of 15 Official leaderboard
Drone-Bench 68.9% #8 of 19 Official leaderboard
ClockBench 32.2% #16 of 37 Official leaderboard
BabyVision 51.6% #12 of 27 Official leaderboard
AgenticVBench 23.8% #4 of 9 Official leaderboard
SpatialWorld Physical 9.20% #4 of 9 Official leaderboard
Omni-InteractScope Explicit Request 18.6% #2 of 4 Official leaderboard
Chartography 26.1% #22 of 43 Official leaderboard
ArtifactArena 1,089 #11 of 21 Official leaderboard
AgentWorldBench 54.6 #7 of 13 Official leaderboard
VoxelBench 1,619 #32 of 56 Official leaderboard
EnigmaEval 36.8% #3 of 5 Official leaderboard
DesignArena Code to Video 1,119 #18 of 29 Official leaderboard
BVB 64.6 #29 of 45 Official leaderboard
MMMU-Pro Extended 31% #25 of 38 Official leaderboard
Blueprint-Bench 2 26.5% #27 of 38 Official leaderboard
CharXiv Extended 76.9% #28 of 38 Official leaderboard
Furniture Assembly Benchmark 26.7% #29 of 35 Official leaderboard
InteractiveBench 20.8% #10 of 12 Official leaderboard
Code2Video Benchmark 1,454 #16 of 19 Official leaderboard
Claw-Eval-MM 45.7% #10 of 11 Official leaderboard
Dual-Arm Sim 8.50 #10 of 11 Official leaderboard
CineRecon 20.1 #12 of 12 Official leaderboard
LongExtractionBench 48.6% #3 of 3 Official leaderboard
Omni-Cloze 57.2% #3 of 3 Official leaderboard

Safety and security

BenchmarkScoreRankSource
ABC Bench Liquid-Handling Robot 100% #1 of 34 Official leaderboard
nocot-bench (NCRI) 115.3 #10 of 185 Official leaderboard
Agentic BAIM-LLM Evaluation (ABLE) 84.6% #1 of 18 Official leaderboard
BinaryAudit 49.1% #2 of 20 Official leaderboard
BrowseSafe++ 89.9% #2 of 15 Official leaderboard
HarnessAudit-Bench 0.41 #1 of 7 Official leaderboard
Human Pathogen Capabilities Test v2 (HPCT v2) 65.4% #11 of 71 Official leaderboard
SMDD-Bench 39% #2 of 7 Official leaderboard
EvalDetectBench 0.87 #3 of 10 Official leaderboard
ABC Bench Fragment Design 94% #9 of 27 Official leaderboard
BioSimBench 0.15 #4 of 12 Official leaderboard
LiveSecBench 58.2 #13 of 34 Official leaderboard
PCT (Polarized Contrastive Pairs) 42.2% #20 of 46 Official leaderboard
LifeSciBench 51.5% #2 of 4 Official leaderboard
ExploitBench 26% #5 of 9 Official leaderboard
KORA 31.6% #25 of 42 Official leaderboard
Dictatorship Eval 53.4% #12 of 20 Official leaderboard
DrugDiscoveryBench 41.9% #12 of 20 Official leaderboard
FORTRESS 29.8% #45 of 73 Official leaderboard
CyberGym 38.8% #22 of 34 Official leaderboard
DistressBench 68.8% #28 of 42 Official leaderboard
BrowseSafe-Adapt 92.9% #4 of 6 Official leaderboard
CyberGym-E2E 43.8% #4 of 5 Official leaderboard
ExploitGym 1.38% #15 of 18 Official leaderboard
GeneBench Pro 3.10% #10 of 12 Official leaderboard
SRE-Bench 0% #32 of 35 Official leaderboard
Suicide Memory Bench 0.43 #11 of 12 Official leaderboard
CyberBench (Vals) 36.4% #26 of 26 Official leaderboard
BiomniBench 44.3 #8 of 8 Official leaderboard
Harmful manipulation efficacy study 3.60 #1 of 1 Lab model card

Other

BenchmarkScoreRankSource
LMCA 53.8% #22 of 147 Official leaderboard
Conceptual Reasoning Index 63.8 #26 of 151 Official leaderboard
StudentBench 15.9 #2 of 11 Official leaderboard
ACCoRD 73.5 #33 of 152 Official leaderboard
CAIS Text Capabilities Index 45.8 #16 of 68 Official leaderboard
CAIS Vision Capabilities Index 62.1 #14 of 53 Official leaderboard
SAGE (Vals) 48.7% #26 of 81 Official leaderboard
TutorBench 53% #11 of 20 Official leaderboard
Vals Multimodal Index 56.1% #20 of 33 Official leaderboard
CAIS Risk Index 56.5 #44 of 67 Official leaderboard
SAGE v1.1 (Vals) 83.2% #23 of 29 Official leaderboard
Vals Science Index 27.4% #23 of 25 Official leaderboard

Its lab's numbers put it at 157.5, 1.3 points above its ECI

The Self-Reported Index fits one score to the 13 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 155.2 to 159.8. How we compute it.

Its lab's numbers next to 8 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
Terminal-Bench 2.0 68.5% Terminal-Bench 2.0 (Terminus-2 harness) 80.2% Leaderboard, TongAgents −12 pts
MCP Atlas 69.2% MCP Atlas 78.2% Leaderboard, high −9.0 pts
SWE-Bench Pro 54.2% SWE-Bench Pro (Public) (Single attempt) 46.1% Leaderboard, thinking, mini-SWE-agent +8.1 pts
SWE-bench Verified 80.6% SWE-Bench Verified (Single attempt) 75.6% Epoch AI run +5.0 pts
Humanity's Last Exam 51.4% Humanity's Last Exam (full set, text + MM), Search (blocklist) + Code; No tools = 44.4% 46.4% Leaderboard, thinking high +5.0 pts
GPQA Diamond 94.3% GPQA Diamond (No tools) 94.4% Epoch AI run, high Same
ARC-AGI-2 77.1% ARC-AGI-2 (ARC Prize Verified) 77.1% Leaderboard Same
MMMU-Pro 80.5% MMMU-Pro (No tools) 80.5% Leaderboard, thinking, high Same

Google DeepMind's models over time

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

50 100 150 2023 2024 2025 2026 2027 Gemini 4 Carbon Gemini 3.8 Flash Gemini 3.5 Flash-Lite Gemini 3 Pro Gemini 2.5 Flash Gemini 2.5 Flash-Lite Gemini 1.5 Pro 50 100 150 2023 2024 2025 2026 2027 Gemini 4 Carbon Gemini 3.8 Flash

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 3.1 Pro model card Feb 19, 2026

Its system card

Every result in the longer documents the lab published with it. 10 of them aren't in the launch post.

Gemini 3.1 Pro Model Card Feb 19, 2026

Capabilities

Safety

Alignment

Bio and chem risk

AI R&D risk

Persuasion