BenchLeader
Menu
Best LLMsChoose a modelCoding plansLeaderboardsModelsBenchmarksProvidersReleases4

Gemini 3.1 Pro vs Kimi K3

Full profiles: Gemini 3.1 Pro (Google) and Kimi K3 (Moonshot AI): every benchmark, price and rank.

Updated · Data from 14 sources

As of 2 Nov 2026, Gemini 3.1 Pro and Kimi K3 are neck and neck on quality (60.6 vs 62.1 on the BenchLeader Index, inside the ±5.1 margin), and Gemini 3.1 Pro is 1.3× cheaper at $4.50 vs $6.00 per 1M tokens blended. Choose Gemini 3.1 Pro for reasoning (66 vs 60) and multimodal (65 vs 63) and fast first answers (17.7 s vs 37.2 s), where it also streams ~2× faster; choose Kimi K3 for human preference (68 vs 58), coding (64 vs 57), and knowledge (60 vs 56).

Key facts (as of 2 Nov 2026)

  • Gemini 3.1 Pro scores 60.6 on the BenchLeader Index and Kimi K3 scores 62.1 (BenchLeader, 2 Nov 2026).
  • Gemini 3.1 Pro costs $4.50 per 1M tokens blended (list price, 2 Nov 2026).
  • Kimi K3 costs $6.00 per 1M tokens blended (list price, 2 Nov 2026).
  • Gemini 3.1 Pro streams ~2× faster than Kimi K3 (Artificial Analysis, 7-day median, 2 Nov 2026).
  • Of the 42 benchmarks both models have, Kimi K3 leads 31 (BenchLeader, 2 Nov 2026).
  • On Arena WebDev, Gemini 3.1 Pro is rated 1447 and Kimi K3 1656 (LMArena, as published by 2 Nov 2026).
Gemini 3.1 Pro benchmarks Google Kimi K3 benchmarks max Moonshot AIAdd a third model

Which to pick for…

The pick's numbers come first. Category scores on the BenchLeader scale (50 is the average model, 15 points one standard deviation). High confidence means a lead of half a standard deviation or more; under one point is called either.

Use casePickWhyConfidence
CodingKimi K364 vs 57 category score; Arena WebDev 1656 vs 1447High
Agents & tool useKimi K359 vs 57 category score; τ³-Banking (AA) 46.0% vs 21.4%Medium
MathsKimi K363 vs 61 category score; FrontierMath Tiers 1–3 72.2% vs 59.6%Medium
ReasoningGemini 3.1 Pro66 vs 60 category score; GPQA Diamond 94.1% vs 93.1%Medium
KnowledgeKimi K360 vs 56 category score; GDP.pdf 19.0% vs 17.0%Medium
Long documentsKimi K369 vs 66 category score; AA-LCR v1.1 88.7% vs 82.0%Medium
Images and other mediaGemini 3.1 Pro65 vs 63 category score; MMMU-Pro 82.4% vs 80.5%Medium
Chat quality (human preference)Kimi K368 vs 58 on Arena votesHigh
Lowest costGemini 3.1 Pro$4.50 vs $6.00 per 1M blended; agentic coding session $0.168 vs $0.240–
Fastest first answerGemini 3.1 Pro17.7 s (Gemini 3.1 Pro) vs 37.2 s (Kimi K3 (max)) to the first answer token–
High volume on a budgetNeither flagshipA smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs–

Key differences

  • Gemini 3.1 Pro and Kimi K3 (max) are neck and neck on quality (60.6 vs 62.1, within the ±5.1 margin).
  • Gemini 3.1 Pro is stronger in instruction following, multimodal, reasoning.
  • Kimi K3 (max) is stronger in agents & tools, coding, human preference, knowledge, long context, maths.
  • Gemini 3.1 Pro is 1.3× cheaper ($4.50 vs $6.00 per 1M blended).
  • Gemini 3.1 Pro streams 2.2× faster (143 vs 66 tokens per second).

At a glance

SpecGemini 3.1 ProKimi K3
DeveloperGoogleMoonshot AI
Released19 Feb 202616 Jul 2026
Context window (tokens)1,048,5761,048,576
Compared atGemini 3.1 ProKimi K3 (max)
Open weightsNoYes
Price in / out / cached input per 1M$2.00 / $12.00 / $0.200$3.00 / $15.00 / $0.300

Category by category

Category scores on the BenchLeader scale: 50 is the average evaluated model, 15 points is one standard deviation. A gap under one point is called level.

CategoryGemini 3.1 ProKimi K3 (max)Leader
Reasoning66.360.0Gemini 3.1 Pro
Coding56.764.4Kimi K3
Agents & tools57.159.2Kimi K3
Maths60.763.2Kimi K3
Knowledge55.959.8Kimi K3
Instruction following66.9–not comparable
Human preference58.468.0Kimi K3
Multimodal64.663.1Gemini 3.1 Pro
Long context65.969.4Kimi K3

On one chart

Each domain on the index's own scale, where 50 is an average model. Click a domain to see the benchmarks behind it. Pick up to 5 models from the list.

2550 avg75ReasoningCodingAgents & toolsMathsKnowledgeLong contextMultimodalHumanpreferenceGemini 3.1 Pro, Reasoning: 66.3Gemini 3.1 Pro, Coding: 56.7Gemini 3.1 Pro, Agents & tools: 57.1Gemini 3.1 Pro, Maths: 60.7Gemini 3.1 Pro, Knowledge: 55.9Gemini 3.1 Pro, Long context: 65.9Gemini 3.1 Pro, Multimodal: 64.6Gemini 3.1 Pro, Human preference: 58.4Kimi K3, Reasoning: 60.0Kimi K3, Coding: 64.4Kimi K3, Agents & tools: 59.2Kimi K3, Maths: 63.2Kimi K3, Knowledge: 59.8Kimi K3, Long context: 69.4Kimi K3, Multimodal: 63.1Kimi K3, Human preference: 68.0

Not drawn, measured for fewer than two of these models: Instruction following.

ModelIndex

Where sources disagree

Across the 42 benchmarks both have, Gemini 3.1 Pro leads 11, Kimi K3 leads 31. Every overall index that rates both puts Kimi K3 ahead.

Overall indexGemini 3.1 ProKimi K3 (max)Leader
BenchLeader Index60.662.1neck and neck
Arena Text14871488Kimi K3

Each publisher runs its own tasks, harnesses and effort settings, so their overall verdicts can differ; BenchLeader reconciles them on one scale. † measured at another effort setting of the same model (the publisher did not test the one compared here).

Price, speed and latency

List prices blended at three input tokens per output token. Per-task prices use the token counts on the methodology page. Cost, tokens and time per benchmark task are measured: Artificial Analysis’ averages over its whole Intelligence Index run.

MeasureGemini 3.1 ProKimi K3 (max)
Blended price per 1M tokens$4.50$6.00
Input / output per 1M$2.00 / $12.00$3.00 / $15.00
Cached input per 1M$0.200$0.300
Price per task: chat reply$0.0044$0.0057
Price per task: summarize a 30-page report$0.031$0.045
Price per task: code edit$0.030$0.041
Price per task: agentic coding session$0.168$0.240
Price per task: structured extraction$0.0064$0.0090
Agentic coding session, input mostly cached$0.087$0.118
Output speed143 tok/s66 tok/s
Time to first answer17.7 s37.2 s
Time to a complete answer21.2 s44.8 s

Each model is compared at its best-ranked reasoning-effort setting. Gemini 3.1 Pro scores 60.6 and takes 18 s to its first answer token; Gemini 3.1 Pro (high) scores 60.3 in 18 s. Kimi K3 (max) scores 62.1 and takes 37 s to its first answer token; Kimi K3 scores 59.1 in 1 s, Kimi K3 (low) scores 53.8 in 40 s. Higher effort means more thinking before the answer: better scores, longer waits and more tokens billed. How reasoning effort changes scores and cost.

Every benchmark

188 benchmarks, grouped by category; each name links to the board and its leaders. † measured at another effort setting of the same model, shown for reference and not counted as a win.

MetricPublisherGemini 3.1 ProKimi K3 (max)
BenchLeader IndexBenchLeader60.662.1
Agents & tools scoreBenchLeader57.159.2
Coding scoreBenchLeader56.764.4
Human preference scoreBenchLeader58.468.0
Instruction following scoreBenchLeader66.9–
Knowledge scoreBenchLeader55.959.8
Long context scoreBenchLeader65.969.4
Maths scoreBenchLeader60.763.2
Multimodal scoreBenchLeader64.663.1
Reasoning scoreBenchLeader66.360.0
Blended price $/MBenchLeader$4.50$6.00
Output speedBenchLeader143 tok/s66 tok/s
Time to first answerBenchLeader17.7 s37.2 s
Context windowBenchLeader1,048,576 tokens1,048,576 tokens
Reasoning
GPQA DiamondEpoch AI Benchmarking Hub94.1%93.1%
Humanity's Last ExamScale AI / CAIS46.4%–
SimpleBenchSimpleBench79.6%60.7%
Arena Hard PromptsArena (LMArena)15081517
LiveBench 2.0 ReasoningLiveBench–90.7%†
LiveBench 1.0 ReasoningLiveBench84.0%†90.7%†
GPQA Diamond (AA)Artificial Analysis94.1%93.5%
Humanity's Last Exam (AA)Artificial Analysis47.0%46.9%
HLE-Diamond (AA)Artificial Analysis31.4%26.9%
GPQA Diamond (Vals)Vals AI95.5%†92.9%†
EnigmaEvalScale AI SEAL36.8%†–
ARC-AGI-1ARC Prize98.0%94.5%
ARC-AGI-2ARC Prize77.1%60.4%
ARC-AGI-3ARC Prize0.4%–
CritPtArtificial Analysis17.7%23.4%
CritPt v2 hard subset (AA)Artificial Analysis–39.5%
MysteryMechanismVals AI20.7%†–
Chess PuzzlesEpoch AI Benchmarking Hub51.0%†36.0%
Mystery Game PuzzlesEpoch AI Benchmarking Hub34.0%†26.0%
CL-benchEpoch AI Benchmarking Hub20.8%–
CL-bench LifeEpoch AI Benchmarking Hub16.9%–
LMCAEpoch AI Benchmarking Hub53.8%†52.7%
DTBenchEpoch AI Benchmarking Hub95.5%†91.2%
ForecastBenchForecastBench64.3%64.7%
Coding
SciCodeSciCode58.7%59.5%
Show all 188 benchmarks
MetricPublisherGemini 3.1 ProKimi K3 (max)
Coding
WeirdMLWeirdML72.1%82.6%
FrontierCodeCognition–44.2%†
GSO-BenchGSO-Bench22.6%–
Arena CodingArena (LMArena)15211540
Arena WebDevArena (LMArena)14471656
LiveBench 2.0 CodingLiveBench–59.3%†
LiveBench 1.0 CodingLiveBench76.5%†81.5%†
LiveCodeBenchVals AI88.5%†87.2%
IOIVals AI51.8%†48.9%
SWE-bench (Vals)Vals AI78.8%†93.4%†
SWE-Bench ProScale AI SEAL46.1%†–
Code MigrationVals AI17.3%†16.1%
Code Migration v1.1Vals AI–28.6%
ProgramBenchVals AI0.0%†2.0%†
SWE-bench MultimodalVals AI–44.2%
Vibe Code Bench 1-100Vals AI6.7%†18.2%
Vibe Code Bench v1.1Vals AI32.0%†85.0%
SWE-Bench Pro (private)Scale AI SEAL46.1%†–
SWE Atlas: Codebase QnAScale AI SEAL13.5%–
SWE Atlas: RefactoringScale AI SEAL33.8%–
SWE Atlas: Test WritingScale AI SEAL29.8%–
MirrorCodeEpoch AI Benchmarking Hub8.9%†–
Surface Evolver BenchEpoch AI Benchmarking Hub–93.0%
DeepSWE v1.1Epoch AI Benchmarking Hub11.7%68.5%
ALE-BenchEpoch AI Benchmarking Hub1160.61524.5
AlgoTuneAlgoTune2.0–
FrontierSWEEpoch AI Benchmarking Hub–25.9%
Terminal-Bench 4.0 (AA)Artificial Analysis4.0%12.6%
Terminal-Bench 2.1 (AA)Artificial Analysis73.8%85.0%
Startrise FrontendStartrise–79.8%†
GameBench 2 OverallDuelLab GameBench 2–48.8†
GameBench 2 by reasoning settingDuelLab GameBench 2–45.1
FrontierCode 1.1 (AA)Artificial Analysis–43.6%
FrontierCode 2.0 (AA)Artificial Analysis–10.9%†
Agents & tools
Terminal-BenchTerminal-Bench80.2%–
APEX-AgentsMercor35.1%47.8%†
Arena AgentArena (LMArena)-4.51.1
LiveBench Long AgenticLiveBench–26.4%†
LiveBench BrowserLiveBench–75.2%†
LiveBench Agentic CodingLiveBench44.1%†62.2%†
Terminal-Bench HardArtificial Analysis53.8%–
τ²-Bench Telecom (AA)Artificial Analysis95.6%–
Terminal-Bench 2.1 (Vals)Vals AI70.8%†80.9%†
MCP AtlasScale AI SEAL78.2%†82.3%
HiL-BenchScale AI SEAL35.3%42.7%†
τ³-bench Bankingτ²-bench26.0%†37.1%
GDPval-AA v2.1, normalised scoreArtificial Analysis14.6%–
GDPval-AA v2.1Artificial Analysis799–
τ³-Banking (AA)Artificial Analysis21.4%46.0%
ITBench SRE (AA)Artificial Analysis30.3%47.7%
Analyst Agent (AA)Artificial Analysis41.3%38.8%
APEX-Agents (AA)Artificial Analysis32.0%41.3%
CyberBenchVals AI–75.2%
Finance Agent v2Vals AI43.0%†53.1%
Finance Agent v2.1Vals AI–60.0%
Harvey's Legal Agent Benchmark v1.1Vals AI–12.9%
Harvey's Legal Agent Benchmark v1Vals AI0.0%†12.9%
Legal Research BenchVals AI20.7%†46.1%
SREBenchVals AI0.0%†–
Tax Agent BenchVals AI9.3%†33.4%
Terminal-Bench 2.0 (Vals)Vals AI67.4%†–
Terminal-Bench 4.0 (Vals)Vals AI2.5%†17.2%
Terminal-Bench ScienceVals AI1.4%†1.4%
Time Horizon Index: KSPVals AI–10.5%
BALROGBALROG57.0%–
DeepResearch BenchDeepResearch Bench47.8%†–
PostTrainBenchEpoch AI Benchmarking Hub22.0%32.0%†
ExploitBenchEpoch AI Benchmarking Hub26.1%–
METR Time HorizonsMETR77.0%–
GBAEvalEpoch AI Benchmarking Hub0.8%48.4%†
Vending-Bench 2Andon Labs911.25165.0†
AutomationBenchArtificial Analysis35.4%58.3%
GDP.pdf (AA)Artificial Analysis17.8%22.0%
Harvey LAB-AA v1.1Harvey0.0%5.3%
EnterpriseOps-GymArtificial Analysis42.2%45.3%
AA-Briefcase v1.1Artificial Analysis4571498
GDPxGDPx by SemiAnalysis–3.7%
Terminal-Bench-Science 0.1 (AA)Artificial Analysis–8.1%
CWE-Bench (AA)Artificial Analysis–53.3%
DeepSec-Bench (AA)Artificial Analysis–12.4%
CyberGym end-to-end (AA)Artificial Analysis–58.0%
BrowseCompOpenAI31.2%–
DeepSearchQAGoogle60.2%–
FACTS SearchGoogle83.7%–
Maths
FrontierMath Tiers 1–3Epoch AI Benchmarking Hub59.6%72.2%
FrontierMath Tier 4Epoch AI Benchmarking Hub26.8%39.0%
OTIS Mock AIMEEpoch AI Benchmarking Hub95.6%97.2%
ProofBenchVals AI26.0%87.0%†
LiveBench MathematicsLiveBench91.0%†84.4%†
AIME (Vals)Vals AI98.1%†–
AIME 2026MathArena98.3%96.7%†
HMMT February 2026MathArena94.7%97.0%†
USAMO 2026MathArena74.4%–
MathArena ApexMathArena60.9%65.6%†
Arena MathsArena (LMArena)14881499
Knowledge
SimpleQA VerifiedEpoch AI Benchmarking Hub73.5%†50.6%
LiveBench 2.0 Data AnalysisLiveBench–78.7%†
LiveBench ResearchLiveBench–79.0%†
LiveBench 1.0 Data AnalysisLiveBench78.5%†78.7%†
AA-Omniscience v1.1Artificial Analysis–-1.7†
MMLU-ProVals AI91.0%†88.0%
LegalBenchVals AI87.4%†86.2%
CorpFinVals AI64.5%†71.6%†
TaxEvalVals AI72.9%†75.7%†
MedQAVals AI96.4%†–
PRBench FinanceScale AI SEAL41.9%–
PRBench LegalScale AI SEAL44.0%–
MultiNRCScale AI SEAL64.7%–
BioMysteryBenchVals AI60.0%†72.6%
CaseLaw v2Vals AI64.8%†–
Excel Modeling BenchmarkVals AI52.6%†66.7%
Excel Modeling Benchmark v1.1Vals AI–71.3%
MedCodeVals AI59.1%†49.4%
MedScribeVals AI76.1%†88.0%
Public Benefits Bench v1.1Vals AI53.8%†68.2%
DrugDiscoveryBenchScale AI SEAL41.9%†–
EBR-benchEpoch AI Benchmarking Hub14.3%–
GDP.pdfEpoch AI Benchmarking Hub17.0%19.0%
AA-Omniscience v1.1: accuracyArtificial Analysis–42.4%†
AA-Omniscience v1.1: non-hallucinationArtificial Analysis–23.4%†
Global-MMLU-Lite (AA)Artificial Analysis93.2%–
HealthBench Professional (AA)OpenAI–53.2%
Instruction following
LiveBench LanguageLiveBench85.4%†85.5%†
LiveBench Instruction FollowingLiveBench79.1%†71.4%†
IFBenchArtificial Analysis77.1%–
MultiChallengeScale AI SEAL71.4%–
TutorBenchScale AI SEAL53.0%–
Arena Instruction FollowingArena (LMArena)14801490
Human preference
Arena TextArena (LMArena)14871488
EQ-Bench 4EQ-Bench11321319†
Arena Creative WritingArena (LMArena)14811463
Arena Multi-turnArena (LMArena)14961502
Arena Longer QueriesArena (LMArena)14991503
Multimodal
Arena VisionArena (LMArena)1296–
LiveBench DesignLiveBench–77.2%†
MMMU-ProArtificial Analysis82.4%80.5%
MMMU-Pro (official)MMMU80.5%†–
MMMU-Pro (Vals)Vals AI88.2%†88.2%†
MortgageTaxVals AI69.4%†66.3%†
SAGEVals AI83.2%†86.8%
VTBScale AI SEAL29.0%–
Blueprint-Bench 2Epoch AI Benchmarking Hub26.5%29.5%†
Code2Video-AAArtificial Analysis–1041
AA-3DArtificial Analysis–982
Long context
AA-LCR v1.1Artificial Analysis82.0%88.7%
Arena DocumentArena (LMArena)1459–
MLCRArtificial Analysis11.1%21.7%
Composite
Epoch Capabilities IndexEpoch AI Benchmarking Hub155.2158.3†
LiveBench 2.0LiveBench–64.6%†
LiveBench 1.0LiveBench77.0%†79.2%†
AA Intelligence Index v5.0Artificial Analysis–17.4†
AA Intelligence Index v4.3.2Artificial Analysis29.743.6
AA Progress Index v1Artificial Analysis296.1319.9
Vals IndexVals AI–53.0
Vals Finance IndexVals AI–58.0%
Vals RSI IndexVals AI–20.9%
Vals Science IndexVals AI27.4%†–
Vals Multimodal IndexVals AI56.1%†73.4%†
Vals Router IndexVals AI–53.0%†
AA Cyber IndexArtificial Analysis–41.3
Safety & honesty
FORTRESSScale AI SEAL29.8%26.6%
MASKScale AI SEAL42.4%–
Metadata
AA Openness IndexArtificial Analysis–38.9

Data as of 2026-11-02. Best configuration of each model; every score links to its source on the model pages.

DuelLab GameBench 2, release gb2-20261026-cb69502a12d9 (https://benchmarks.duellab.org/releases/gb2-20261026-cb69502a12d9/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Scores mapped to BenchLeader model names; not endorsed by DuelLab.

What changed

  • Kimi K3 (max): 1 new result (MirrorCode).
  • Kimi K3 (max): 1 score revision.
  • Gemini 3.1 Pro: BenchLeader Index 61.6 → 60.5.
  • Kimi K3 (max): BenchLeader Index 63.1 → 62.1.
  • Gemini 3.1 Pro: 2 new results (AA Progress Index v1, HLE-Diamond (AA)).
  • Kimi K3 (max): 3 new results (AA Progress Index v1, HLE-Diamond (AA), CritPt v2 hard subset (AA)); 1 score revision.
  • Gemini 3.1 Pro: BenchLeader Index 59.3 → 61.6.
  • Gemini 3.1 Pro: Gemini 3.1 Pro cost per task moved from $1.30/M to $1.44/M (11% up).
  • Gemini 3.1 Pro: 2 score revisions.
  • Kimi K3 (max): 7 new results (Code Migration v1.1, Excel Modeling Benchmark v1.1, Finance Agent v2.1…).

Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs

Gemini 3.1 Pro vs Kimi K3: questions

Is Gemini 3.1 Pro better than Kimi K3?
Gemini 3.1 Pro and Kimi K3 (max) are neck and neck on quality (60.6 vs 62.1, within the ±5.1 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 2 Nov 2026, so check the category scores for your use.
Gemini 3.1 Pro vs Kimi K3: which is cheaper, and will that change?
Gemini 3.1 Pro is cheaper today: $4.50 against $6.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
Is Gemini 3.1 Pro or Kimi K3 better for coding?
Kimi K3 scores higher in coding (64 vs 57 on the category index, where 50 is average). On Arena WebDev, the top shared coding benchmark, it is 1447 for Gemini 3.1 Pro against 1656 for Kimi K3.
Which is faster, Gemini 3.1 Pro or Kimi K3?
Gemini 3.1 Pro streams ~2× faster (143 vs 66 output tokens per second, Artificial Analysis 7-day median).
Is Gemini 3.1 Pro or Kimi K3 better for agents and tool use?
Kimi K3 scores higher in agentic tasks (59 vs 57 on the category index, where 50 is average). On τ³-Banking (AA), the top shared agents & tools benchmark, it is 21.4% for Gemini 3.1 Pro against 46.0% for Kimi K3.
More questions (4)
Which is better for long documents?
Kimi K3 scores higher on long-context benchmarks (69 vs 66). Both accept 1,048,576 tokens of context.
Which has lower latency?
Gemini 3.1 Pro answers first: 17.7 s against 37.2 s to the first answer token, as Artificial Analysis measures it. Reasoning effort drives most of this: a higher setting thinks longer before it answers.
Do independent benchmarks agree on Gemini 3.1 Pro vs Kimi K3?
Across the 42 benchmarks both have, Gemini 3.1 Pro leads 11, Kimi K3 leads 31. Every overall index that rates both puts Kimi K3 ahead.
When was Kimi K3 released?
Kimi K3 was released on 16 Jul 2026.