Atlas

Compare Models

Gemini 3.7 Flash vs Kimi K3.1 Pro

Kimi K3.1 Pro leads by 4.5 capability points (≈95% interval +2.3 to +6.6): a clear lead over Gemini 3.7 Flash · P(higher) ≈ 100%

Wins 59 of 82 shared benchmarks (20 lost, 3 tied)

Capability vs. relative cost
Gemini 3.7 FlashKimi K3.1 ProClosed weightsOpen weights

Each line runs through one model’s priced reasoning settings, cheapest to most expensive; grey points are other models’ priced settings in the same range. Results whose reasoning effort the source didn’t report are left out wherever a model has named effort levels. Capability and relative cost are the published estimates (relative cost is cost per task as a multiple of GPT-5 (high)); hover a setting, or use the arrow keys, to set it beside the other model’s same setting.

Gemini 3.7 FlashvsKimi K3.1 Procapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins20 – 59 · 82 shared, 3 tied

Comparing matching setups only (23 benchmarks excluded) · suites counted once
← Gemini 3.7 FlashKimi K3.1 Pro →

Gemini 3.7 FlashvsKimi K3.1 Prowins by categoryData Nov 15, 2026Data: Arena, Vals AI, SecureBio and 20 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Kimi K3.1 Pro minus Gemini 3.7 Flash.

Terminal-Bench 4.0Code6.165.7+59.6
ProofBench v1.1Math58.091.0+33.0
ZeroBench (Main questions, pass@5, with tools)Multimodal30.063.0+33.0
AA-Briefcase v1.1 EloAgents11151678+30.9
Concept Synth INDUCTION Challenge64 Round 1General QA17.246.9+29.7
FrontierMath Tier 4 (v2) - PrivateMath36.665.9+29.3
Furniture Assembly BenchmarkMultimodal26.755.8+29.2
VoxelBench Text RatingGames18442216+26.5
Terminal-Bench Science 0.1 (Vals)Science7.131.4+24.3
Legal Research Bench HealthLegal50.073.3+23.3
Concept Synth INDUCTION Challenge100 Round 1General QA11.034.0+23.0
Creative Writing v3 EloChat & Writing17202173+22.9
Arena Code WebDev — Brand & MarketingCode15531704+21.9
Chess Puzzles (task version 2.0)Games66.045.0−21.0
GDPval-AA v2.1Professional Work14041846+19.5
Legal Research Bench ImmigrationLegal36.418.2−18.2
Arena Code WebDev — ReactCode15841700+15.6
Vibe Code Bench v1.1Code70.484.9+14.5
Arena Code WebDev — SimulationsCode16181745+14.4
Arena Code WebDev — FrontendCode15921704+14.3
IOI (Vals v2)Code67.881.9+14.1
MineBench — Bradley-Terry RatingGames17651992+13.7
Legal Research Bench Administrative/RegulatoryLegal42.456.1+13.6
Arena Code WebDev — HTMLCode15931689+13.4
LMArena Agent Arena Net Improvement (Steering Burden)Agents-2.24.9+13.4
Bug Hunt BenchCode21.034.3+13.3
NoCoT-Bench NCRI 15.2General QA126103−12.8
Legal Research Bench Weighted PassLegal84.171.4−12.7
DAYJOB: Finance (October 2026 rescoring)Finance & Accounting5.317.0+11.7
AutomationBench-AA 1.0.6Agents62.073.3+11.3
EnterpriseBench: CoreCraftProfessional Work55.966.1+10.2
Arena Code WebDev — Reference-Based DesignCode16081684+10.0
DeepSWE v1.1Code65.575.5+10.0
Arena Code WebDev — GamingCode16381724+9.6
LiveBench 2.0Agents67.276.7+9.5
Legal Research Bench Business/CommercialLegal37.227.9−9.3
Arena Code WebDev — Content Creation ToolsCode15741632+9.1
ALE-BenchCode18762132+9.0
LAB-Bench ProtocolQA (SecureBio)Science86.878.1−8.7
Humanity's Last Exam DiamondGeneral QA31.239.4+8.2
Design Arena Overall Code CategoriesCode13041353+8.1
LAB-Bench CloningScenarios (SecureBio)Science99.491.5−7.9
LMCAGeneral QA50.457.6+7.2
Creative Writing v3 Rubric ScoreChat & Writing15.216.6+7.0
LMArena Agent Arena Steering BurdenAgents-7.36.7+7.0
SimpleQA Verified (Epoch anti-abstention)General QA69.262.2−7.0
FrontierMath Tiers 1-3 (v2) - PrivateMath71.678.2+6.7
WCB (SecureBio)Science58.051.5−6.5
Tax Agent BenchFinance & Accounting57.763.9+6.2
MBCT (SecureBio)Science55.060.7+5.7
GBENCH One-shot Coding GScoreGames52.658.2+5.6
Humanity's Last ExamGeneral QA45.651.1+5.5
LMArena Agent Arena Praise vs ComplaintAgents-2.98.0+5.4
Legal Research Bench CriminalLegal33.328.1−5.3
APEX-Accounting 1.1Finance & Accounting57.262.4+5.2
Mystery Game PuzzlesGames37.042.0+5.0
CharXiv Reasoning (with tools)Multimodal88.793.4+4.7
Conceptual Reasoning IndexIndexes64.667.8+4.5
Arena Code WebDev — Data & AnalyticsCode15771606+4.3
Arena Code WebDev — Consumer ProductCode15741607+4.2
APEX-Agents 1.1Professional Work79.475.9−3.5
DTBenchGeneral QA96.893.3−3.5
NoCoT-Bench NCKI (kspine v3)General QA121116−3.3
LMArena Agent Arena Confirmed SuccessAgents-3.52.7+3.1
BuseyBench SVGMultimodal6.66.9+3.0
Epoch Capabilities Index (ECI)Indexes158162+2.9
OTIS Mock AIME 2024-2025Math97.2100.0+2.8
VCT v2 (SecureBio)Science48.646.2−2.4
HPCT (SecureBio)Science64.763.5−1.2
LMArena Text Arena EloChat & Writing14891483−1.1
LAB-Bench LitQA2 (SecureBio)Science54.253.1−1.1
MMMU ProMultimodal84.984.1−0.8
Bio Capabilities Index (SecureBio)Indexes162161−0.6
GPQA-Bio (SecureBio)Science81.682.2+0.6
ACCoRDGeneral QA77.077.5+0.5
LMArena Agent Arena Bash ReliabilityAgents6.27.2+0.5
MMLU-Bio (SecureBio)Science91.491.2−0.2
LMArena Agent Arena Tool HallucinationAgents0.71.0+0.2
VCT (SecureBio)Science49.349.4+0.1
Legal Research Bench Civil LitigationLegal29.629.60.0
Legal Research Bench Constitutional/Civil RightsLegal31.631.60.0
Legal Research Bench FamilyLegal13.613.60.0

Gemini 3.7 FlashvsKimi K3.1 Pronormalised score marginsData Nov 15, 2026Data: Arena, Vals AI, SecureBio and 20 othersVillage LLM Atlasatlas.village.dev

Win counts use each benchmark’s declared score direction. Represented suite slices and the derived Capability score are excluded from headline totals. Missing or incompatible setups are excluded, never counted as losses.
Loading Atlas data…