Kimi K3

Kimi K3 ranks first on 6 of the 386 benchmarks with five or more models scored.

Moonshot AI Released July 16th, 2026 Family Kimi K3 The lab's model card Compare with others

411

Benchmarks with a score

Ranked first, out of five or more models

In the top three

Moonshot AI Jul 2026

Kimi K3

Open weights Flagship

158.6 #24 of 256

Epoch Capabilities Index

Self-Reported Index 159.5 #35 of 404

MMMU-Pro 81.6% #10 /90
DeepSWE 68.5% #8 /28
AutomationBench 22.7% #14 /42

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
Creative Writing v3	81.8	#4 of 120	Official leaderboard
Joke Arena	1,573	#1 of 25	Official leaderboard
LMArena (Elo)	1,487	#16 of 335	Official leaderboard
AA-LCR (Artificial Analysis Long Context Reasoning)	74.7%	#3 of 38	Lab model card
EQ-Bench 4	1,319	#4 of 32	Official leaderboard
Extended NYT Connections	93.6%	#13 of 103	Official leaderboard
BullshitBench V2	74%	#17 of 129	Official leaderboard
Humanity's Last Exam	56%	#22 of 154	Lab model card
Longform Creative Writing (EQ-Bench)	79.6	#17 of 119	Official leaderboard
Antidote	1,077	#6 of 36	Official leaderboard
Hemingway-bench	1,069	#8 of 42	Official leaderboard
BullshitBench	58.2%	#21 of 107	Official leaderboard
Lech Mazur Creative Writing	2.35	#10 of 50	Official leaderboard
AgentIF (2026-10 setting)	64.9%	#5 of 24	Official leaderboard
SimpleQA Verified	50.6%	#30 of 94	Epoch AI run
AA-LCR Extended	69.1%	#16 of 49	Official leaderboard
StudyArena Text	1,244	#28 of 85	Official leaderboard
Last Translation Benchmark	11.8	#15 of 36	Official leaderboard
HLE Extended	35%	#15 of 31	Official leaderboard
HANDBOOK.md	28.6%	#10 of 16	Official leaderboard
OpenAI MRCR v2	34.8%	#30 of 47	Official leaderboard
Opening Lines Arena	1,450	#10 of 15	Official leaderboard
ComplexConstraints	51.1%	#11 of 15	Official leaderboard
HLE-Diamond	28.3%	#16 of 20	Official leaderboard
Chat All-in-One Bench	85.2%	#1 of 1	Lab model card
CLIF Bench	52.4%	#1 of 1	Lab model card
Faithfulness (Moonshot)	85.5%	#1 of 1	Lab model card

Math and science

Benchmark	Score	Rank	Source
MathArena Apex Shortlist	92%	#3 of 38	Official leaderboard
Surface Evolver Bench	95%	#2 of 25	Official leaderboard
ProofBench	70%	#5 of 59	Official leaderboard
MathArena (Apex, IMO, USAMO)	65.6%	#4 of 47	Official leaderboard
GPQA Diamond	93.1%	#24 of 229	Epoch AI run
HMMT 2026	97%	#4 of 31	Official leaderboard
CritPt	23.4%	#2 of 13	Lab model card
OTIS Mock AIME 2024-2025	97.2%	#39 of 212	Epoch AI run
DTBench	91.2%	#40 of 180	Official leaderboard
ARC-AGI	94.5%	#22 of 95	Official leaderboard
ProofBench v1.1	87%	#13 of 52	Official leaderboard
FrontierChallenge	17.5%	#3 of 11	Official leaderboard
NatureBench	14.4%	#5 of 18	Official leaderboard
ExplorationBench	79%	#3 of 10	Official leaderboard
FrontierMath Tiers 1-3 v2	72.2%	#26 of 86	Official leaderboard
ForecastBench	0.13	#28 of 82	Official leaderboard
BixBench3	47.0	#7 of 20	Official leaderboard
AIME 2026	96.7%	#11 of 31	Official leaderboard
BioMysteryBench	72.6%	#11 of 30	Official leaderboard
SimpleBench	60.7%	#40 of 106	Official leaderboard
ARC-AGI-2	60.4%	#37 of 98	Official leaderboard
FrontierMath Tier 4 v2	39%	#27 of 68	Official leaderboard
PhysClosureBench	5.17	#3 of 7	Official leaderboard
Into the Unknown	0%	#4 of 9	Official leaderboard
xbench-ScienceQA	71.2%	#13 of 29	Official leaderboard
PaperBenchX	7.53%	#5 of 10	Official leaderboard
Apodex Discovery	2.61	#4 of 8	Official leaderboard
ASI-Bench	37.1	#6 of 11	Official leaderboard
FrontierMath Erdős	0%	#11 of 15	Official leaderboard
HorizonMath	2.70%	#6 of 8	Official leaderboard
Terminal-Bench-Science	7.14%	#13 of 17	Official leaderboard
FrontierPhysics	2.40%	#17 of 22	Official leaderboard
LatchBio Capabilities v1.0	36.4%	#21 of 27	Official leaderboard
Prophet Arena	85.3	#19 of 24	Official leaderboard
Riemann-bench	36.8%	#17 of 20	Official leaderboard
scBench (LatchBio SingleCellBench)	47.5%	#23 of 26	Official leaderboard
EurekaBench	12.9	#7 of 7	Official leaderboard
First Proof Third Batch	-0.86	#3 of 3	Official leaderboard

Coding

Benchmark	Score	Rank	Source
DesignArena Data Visualization	1,351	#3 of 118	Official leaderboard
DesignArena Website	1,337	#5 of 151	Official leaderboard
DesignArena Code	1,352	#8 of 143	Official leaderboard
DesignArena Image to HTML	1,269	#4 of 51	Official leaderboard
SciCode	58.7%	#4 of 49	Lab model card
DesignArena Game Dev	1,353	#12 of 143	Official leaderboard
WebDev Arena	1,674	#9 of 105	arena.ai, via Epoch AI
DesignArena UI Component	1,339	#12 of 139	Official leaderboard
WeirdML	82.6%	#10 of 112	Official leaderboard
DesignArena SVG	1,298	#10 of 103	Official leaderboard
DesignArena Full Stack	1,295	#5 of 49	Official leaderboard
DesignArena 3D Design	1,367	#14 of 136	Official leaderboard
DesignArena Video to Website	1,241	#2 of 19	Official leaderboard
DesignArena React Native	1,256	#7 of 56	Official leaderboard
AndroidBench	90.2%	#4 of 32	Official leaderboard
DesignArena Android (Kotlin)	1,255	#6 of 46	Official leaderboard
Code Arena	1,658	#17 of 128	Official leaderboard
SWE-bench Verified Extended	78.7%	#6 of 44	Official leaderboard
DesignArena Frontend Web App	1,279	#8 of 53	Official leaderboard
Vibe Code Bench	85%	#19 of 120	Official leaderboard
ALE-Bench	1,991	#22 of 126	Official leaderboard
DesignArena Agentic Game Dev	1,203	#5 of 27	Official leaderboard
FrontierSWE v1	81.2%	#1 of 5	Lab model card
DesignArena Agentic Overall	1,254	#16 of 74	Official leaderboard
SWE-Bench Pro V2	97.7%	#4 of 18	Official leaderboard
DesignArena Image to Frontend	1,249	#7 of 31	Official leaderboard
SWE-Together	63.8%	#6 of 22	Official leaderboard
DeepSWE	68.5%	#8 of 28	Official leaderboard
FrontierMemory	71.1%	#2 of 7	Official leaderboard
APEX-SWE	48%	#19 of 65	Official leaderboard
Vision2Web	79%	#3 of 10	Official leaderboard
SWE Atlas Codebase QnA Extended	24%	#12 of 38	Official leaderboard
MLS-Bench	48.3%	#5 of 15	Official leaderboard
TestPrism	24.7%	#2 of 6	Official leaderboard
SWE-Marathon v1.1	48.1%	#7 of 20	Official leaderboard
Landing Page Arena	1,560	#5 of 14	Official leaderboard
Tencent WorkBuddy Bench	76.7%	#5 of 14	Official leaderboard
MirrorCode	54.4%	#14 of 39	Epoch AI run
Image-to-WebDev Arena	1,586	#20 of 55	Official leaderboard
SWE Refactor Bench	19.5%	#3 of 8	Official leaderboard
SWE-Bench Pro V2 Hard	88.2%	#7 of 18	Official leaderboard
GameCraft-Bench	57%	#6 of 15	Official leaderboard
GameDevBench	58%	#7 of 17	Official leaderboard
GBAEval	48.4%	#9 of 21	Official leaderboard
PostTrainBench 1.1	32%	#5 of 11	Official leaderboard
ReactBench	33.3%	#13 of 28	Official leaderboard
TasteVal	45.3×	#10 of 20	Official leaderboard
SWE-bench Science	35.3%	#5 of 10	Official leaderboard
LibraryDesignBench	44.0	#4 of 8	Official leaderboard
Brokk Power Ranking	8.30	#10 of 19	Official leaderboard
DAXBench v2	78%	#25 of 47	Official leaderboard
SWE Atlas v1.1	61%	#15 of 28	Official leaderboard
PostTrainBench 1.2	31.5%	#12 of 21	Official leaderboard
Vibe Code Bench 1-100	18.2%	#15 of 26	Official leaderboard
Doc2Harness	38.3%	#11 of 19	Official leaderboard
FrontierCS 2	20.8%	#11 of 19	Official leaderboard
DesignArena Godot Game Dev	1,205	#14 of 24	Official leaderboard
IOI v2 (Vals)	48.9%	#27 of 45	Official leaderboard
FrontierSWE	25.9%	#15 of 25	Official leaderboard
RSI Arena	982	#6 of 10	Official leaderboard
RSI-Exam	38.2%	#11 of 18	Official leaderboard
MacroscopeBench	72.9%	#10 of 16	Official leaderboard
Vals RSI Index	0.21	#17 of 27	Official leaderboard
SWE-Marathon	42%	#7 of 11	Lab model card
Next.js Agent Evals	86%	#9 of 14	Official leaderboard
DesignArena Blender	1,156	#8 of 12	Official leaderboard
ViBench 2.0	0%	#6 of 9	Official leaderboard
Kimi Code Bench v2	72.9%	#2 of 3	Lab model card
Snorkel Agentic Coding 2.0	26.1%	#11 of 16	Official leaderboard
WeirdML v3	7.54%	#19 of 27	Official leaderboard
Code Migration	16.1%	#55 of 78	Official leaderboard
PostTrainBench	36.6%	#5 of 7	Lab model card
Senior SWE-Bench	20.2%	#17 of 23	Official leaderboard
SWE-bench CLI	4.10%	#12 of 16	Official leaderboard
DeepRSI	820	#6 of 8	Official leaderboard
MiMo Code Bench	60.1%	#6 of 8	Official leaderboard
RSI Bench	32.8	#6 of 8	Official leaderboard
MiMo Visual Coding	70.3%	#7 of 9	Official leaderboard
FrontierCode 2.0	18.9%	#25 of 32	Official leaderboard
Code Migration v1.1	28.6%	#24 of 30	Official leaderboard
SWE-sweep	0.60%	#8 of 10	Official leaderboard
StepCodeBench	43.9%	#4 of 5	Official leaderboard
τ^τ-bench	17.9%	#4 of 5	Official leaderboard
ScarfBench	61.3%	#9 of 11	Official leaderboard
Android Bench 2.0	12%	#14 of 17	Official leaderboard
sudo L7	12.2%	#15 of 18	Official leaderboard
AutoResearchExam	0.43	#10 of 12	Official leaderboard
Legacy-Bench	44%	#5 of 6	Official leaderboard
SWE-Clar	49%	#8 of 9	Official leaderboard
DAYJOB: Data Science	1.30%	#14 of 15	Official leaderboard
OpenDesign Arena	65.7	#21 of 21	Official leaderboard
Coding Experience (Moonshot)	59.9%	#1 of 1	Lab model card
Kimi Webdev Bench	58.6%	#1 of 1	Lab model card

Agents and tools

Benchmark	Score	Rank	Source
ClawProBench	78.4%	#2 of 49	Official leaderboard
DeepSearchQA	95%	#1 of 23	Lab model card
CEO-Bench	$22,148,357	#1 of 19	Official leaderboard
OSWorld-Verified	84.8%	#3 of 41	Lab model card
Toolathlon-Verified	76.5%	#2 of 25	Official leaderboard
BrowseComp	91.2%	#7 of 81	Lab model card
LisanBench	10520.7	#13 of 133	Official leaderboard
MCPMark	94.5%	#1 of 10	Lab model card
MCPMark Verified	96.1%	#1 of 8	Official leaderboard
GAIA	80.4%	#9 of 64	Official leaderboard
Terminal-Bench 2.1	88.3%	#9 of 56	Lab model card
GameHorizon	74.5	#6 of 37	Official leaderboard
Long-Horizon-Terminal-Bench 2.0	56.8%	#3 of 15	Official leaderboard
WitnessBench	18.8%	#3 of 12	Official leaderboard
MCP Atlas	82.3%	#11 of 37	Official leaderboard
WildClawBench	54.5%	#10 of 31	Official leaderboard
AutomationBench	22.7%	#14 of 42	Official leaderboard
ResearchRubrics	76.2%	#2 of 6	Lab model card
BabaIsBench	96%	#11 of 32	Official leaderboard
CLASHarness	21.8%	#5 of 14	Official leaderboard
o11y-bench	81.3%	#9 of 22	Official leaderboard
BrowseComp Extended	28.5%	#18 of 42	Official leaderboard
DiG-bench	22.9%	#3 of 7	Official leaderboard
MobileWorld	43.3%	#8 of 18	Official leaderboard
DuetBench	1,011	#4 of 9	Official leaderboard
Chess Puzzles	37%	#42 of 93	Epoch AI run
Mystery Game Puzzles	26%	#39 of 84	Epoch AI run
Hermes Bench	71.5%	#8 of 17	Official leaderboard
Long-Horizon-Terminal-Bench (LHTB)	37.8%	#12 of 25	Official leaderboard
τ³-bench Banking	37.1%	#12 of 25	Official leaderboard
Agent Arena	0.55%	#28 of 56	Official leaderboard
ApprenticeBench	18%	#9 of 18	Official leaderboard
Swarm Bench	76.3%	#1 of 2	Lab model card
OSWorld-Science	30.2%	#6 of 11	Official leaderboard
Terminal-Bench 2.1 Extended	31%	#25 of 44	Official leaderboard
Terminal-Bench 4.0	22.4%	#23 of 40	Official leaderboard
Terminal-Bench 4.1	23%	#25 of 43	Official leaderboard
Bench on the Clocktower	49.3%	#6 of 10	Official leaderboard
RecreationBench	31.4%	#6 of 10	Official leaderboard
OSWorld 2.0	58.3%	#12 of 19	Lab model card
Learn2Play Bench	68.5	#7 of 11	Official leaderboard
Collaborative Intelligence Benchmark	45.9%	#12 of 18	Official leaderboard
TraceTrial-Bench	74.4%	#6 of 9	Official leaderboard
Time Horizon Index: KSP	10.5%	#9 of 13	Official leaderboard
Web Context (Nimble)	69.4%	#8 of 11	Official leaderboard
SaaS-Bench	24.5%	#3 of 4	Official leaderboard
Cresta-Bench	34.9%	#11 of 14	Official leaderboard
RuneBench	3.75	#61 of 76	Official leaderboard
Realtime-Bench	13.3%	#7 of 8	Official leaderboard
tau3-bench	33.4%	#13 of 14	Lab model card
Agent Behavior Bench	65%	#1 of 1	Lab model card
Deep Research Bench (Moonshot)	90%	#1 of 1	Lab model card
KAET (Kimi Autonomous Execution Tasks)	83.5%	#1 of 1	Lab model card

Professional work

Benchmark	Score	Rank	Source
CorpFin v2	71.6%	#3 of 119	Official leaderboard
TaxEval v2	75.7%	#12 of 129	Official leaderboard
ProfBench	57.9%	#7 of 72	Official leaderboard
LegalBench	86.2%	#16 of 138	Official leaderboard
DesignArena PDF	1,278	#5 of 37	Official leaderboard
Harvey LAB (Legal Agent Benchmark)	12.9%	#12 of 80	Official leaderboard
MedScribe	88%	#15 of 97	Official leaderboard
YC-Bench	$2,053,296	#10 of 59	Official leaderboard
PolicyBench	90.5%	#9 of 50	Official leaderboard
DesignArena Python-PPTX Slides	1,221	#7 of 36	Official leaderboard
Legal Research Bench	46.1%	#20 of 93	Official leaderboard
$OneMillion-Bench	35%	#3 of 13	Official leaderboard
Tax Agent Bench	33.4%	#20 of 84	Official leaderboard
MedCode	49.4%	#23 of 96	Official leaderboard
Public Benefits Bench	68.2%	#12 of 50	Official leaderboard
DesignArena HTML Slides	1,152	#7 of 28	Official leaderboard
JobBench	54.3%	#4 of 16	Official leaderboard
APEX-Agents	41%	#3 of 12	Lab model card
APEX-1	68.4%	#18 of 71	Official leaderboard
CHI-Bench	25.3%	#6 of 23	Official leaderboard
Excel Modeling Benchmark	66.7%	#21 of 76	Official leaderboard
GDPval Extended	2.10%	#12 of 43	Official leaderboard
Agents' Last Exam	28.3%	#9 of 31	Official leaderboard
GDPval-AA v2	1,686	#8 of 27	Lab model card
SnorkelRevOps	13.1%	#5 of 16	Official leaderboard
SnorkelUnderwrite 2.0	27.9%	#5 of 16	Official leaderboard
Immigration Bench	68.3%	#8 of 25	Official leaderboard
Vals Index v2	57.8%	#22 of 65	Official leaderboard
Legora BAR	102.0	#6 of 17	Official leaderboard
MortgageTax	66.3%	#32 of 87	Official leaderboard
Vals Legal Index	71.7%	#14 of 38	Official leaderboard
SnorkelFinance 2.0	17.3%	#6 of 16	Official leaderboard
SnorkelLegal	36.1%	#6 of 16	Official leaderboard
Excel Modeling Benchmark v1.1	71.3%	#16 of 42	Official leaderboard
prinzbench	47.0	#10 of 26	Official leaderboard
Harvey LAB v1.1 (HLAB-Vals-1.1)	12.9%	#16 of 38	Official leaderboard
E-Commerce Bench	2.65	#8 of 19	Official leaderboard
Vals Finance Index	58%	#17 of 40	Official leaderboard
Vals Finance Agent v2	53.1%	#35 of 80	Official leaderboard
Thrive Holdings helpdesk ticket resolution benchmark	48.1%	#7 of 15	Official leaderboard
TRLawBench	9.10	#16 of 34	Official leaderboard
Bedside Bench	89.7%	#8 of 17	Official leaderboard
Vending-Bench 2	$5,165	#33 of 67	Official leaderboard
Genspark Slides Benchmark	72.6%	#5 of 10	Official leaderboard
E-Bench (Tencent)	73.8%	#4 of 8	Official leaderboard
E-Bench-Code (Tencent)	77.6%	#4 of 8	Official leaderboard
Ivo-micro1 ContractBench	57.1%	#4 of 8	Official leaderboard
SheetBench	59.3%	#4 of 8	Official leaderboard
FinStepBench CorporateValuation	60.6%	#3 of 6	Official leaderboard
Harvey LAB-AA v1.0	94.6%	#1 of 2	Lab model card
FoodTruck Bench	$22,404	#30 of 59	Official leaderboard
DAYJOB: Healthcare	4.00%	#19 of 37	Official leaderboard
Holdings-IndividualTaxBench	72.6%	#7 of 13	Official leaderboard
Tuesday Work Index	47.3	#24 of 44	Official leaderboard
ATLAS-Finance	3.67%	#6 of 11	Official leaderboard
ATLAS-Finance v1.1.0	3.67%	#6 of 11	Official leaderboard
EnterpriseBench: CoreCraft	43.1%	#28 of 51	Official leaderboard
TaxCalcBench v2	42%	#18 of 32	Official leaderboard
AA-Briefcase	1,548	#4 of 7	Lab model card
FrontierFinance	46.4%	#10 of 17	Official leaderboard
DAYJOB: Finance	5.50%	#23 of 39	Official leaderboard
DesignArena Spreadsheet	1,116	#21 of 35	Official leaderboard
Hebbia Financial Services Benchmark	60.3%	#6 of 10	Official leaderboard
APEX-Accounting	56.7%	#30 of 49	Official leaderboard
Vals Index v2.1	50.3%	#30 of 49	Official leaderboard
SnorkelManufacturing	10.2%	#10 of 16	Official leaderboard
APEX-Agents 1.1	62.2%	#43 of 65	Official leaderboard
LifeSci Diamond	1.80%	#12 of 18	Official leaderboard
LEXam-hard	29.5%	#10 of 15	Official leaderboard
VWE-Bench	44.5%	#4 of 6	Official leaderboard
SpreadsheetBench 2	34.8%	#2 of 3	Lab model card
Vending-Bench Arena	$3,194	#22 of 32	Official leaderboard
Digital Work Index	16.8	#11 of 16	Official leaderboard
Vals Finance Agent v2.1	60%	#28 of 40	Official leaderboard
Investor's Last Exam	58.2	#7 of 10	Official leaderboard
Agents' Last Exam v1.1	35.3%	#10 of 14	Official leaderboard
WorkplaceAgents	13%	#10 of 14	Official leaderboard
Vals Index v2.2	53%	#29 of 37	Official leaderboard
OfficeQA	63.3%	#11 of 14	Lab model card
GDP.xlsx	13.4%	#16 of 20	Official leaderboard
GDPx	3.70%	#16 of 20	Official leaderboard
Build-A-Chip-Bench	0.20%	#4 of 5	Official leaderboard
Parametric CAD Bench	24.9%	#9 of 11	Official leaderboard
DAYJOB: Accounting	6.00%	#14 of 17	Official leaderboard
Epoch Automation Reports	52.4%	#5 of 6	Official leaderboard
FinStepBench DeepResearch	48.9%	#5 of 6	Official leaderboard
Clipping Bench	$343	#16 of 19	Official leaderboard
BlenderWorld	46.5	#11 of 13	Official leaderboard
MedXpertQA MM Extended	47.7%	#35 of 41	Official leaderboard
EngiWorld	24.3%	#6 of 7	Official leaderboard
DAYJOB: Tax	0%	#13 of 15	Official leaderboard
Quant-Arena	38.3	#8 of 9	Official leaderboard
HW-Bench	7.00%	#14 of 15	Official leaderboard
DAYJOB: Law	4.20%	#16 of 16	Official leaderboard
FinStepBench LiveSearch	70.9%	#6 of 6	Official leaderboard
MBABench 2.0	0%	#5 of 5	Official leaderboard
Business Operation (Moonshot)	8.90%	#2 of 2	Official leaderboard
24/7 ClawBench 2.0	48.3%	#1 of 1	Lab model card
DECK Bench	73.5%	#1 of 1	Lab model card
Finance Bench (Moonshot)	62.6%	#1 of 1	Lab model card
MIRA Bench	64.1%	#1 of 1	Lab model card
Online Experience (Moonshot)	77.9%	#1 of 1	Lab model card
Vals Index v1	74.7%	#1 of 1	Lab model card

Vision, audio and robotics

Benchmark	Score	Rank	Source
Video-MME	90%	#2 of 51	Lab model card
MMVU	82.1%	#2 of 26	Lab model card
DesignArena Code to Video	1,270	#3 of 29	Official leaderboard
MMMU-Pro	81.6%	#10 of 90	Lab model card
PerceptionBench (Moonshot)	58.5%	#2 of 16	Official leaderboard
Math Kangaroo 2025 (MathArena)	91.3%	#3 of 22	Official leaderboard
MATH-Vision (MathVision)	94.3%	#7 of 51	Lab model card
BabyVision	85.7%	#4 of 27	Official leaderboard
ZeroBench	41%	#12 of 76	Official leaderboard
CineRecon	52.4	#3 of 12	Official leaderboard
MDPBench	83.6	#3 of 12	Official leaderboard
CharXiv	84.8%	#20 of 75	Lab model card
VoxelBench	1,907	#16 of 56	Official leaderboard
MineBench	1,697	#26 of 75	Official leaderboard
ArtifactArena	1,164	#8 of 21	Official leaderboard
MMMU-Pro Extended	36%	#16 of 38	Official leaderboard
Chartography	26.6%	#21 of 43	Official leaderboard
VibeVideoBench	3.44	#5 of 9	Official leaderboard
WorldVQA	51%	#5 of 9	Lab model card
Code2Video Benchmark	1,481	#11 of 19	Official leaderboard
ExtractBench	83.2%	#16 of 27	Official leaderboard
RobotWorld	2.38%	#3 of 5	Official leaderboard
Blueprint-Bench 2	29.5%	#24 of 38	Official leaderboard
CharXiv Extended	78.7%	#24 of 38	Official leaderboard
Frontier3DBench	49.2	#7 of 11	Official leaderboard
Furniture Assembly Benchmark	34.2%	#23 of 35	Official leaderboard
Code2Video-Compass	9.87	#10 of 15	Official leaderboard
InteractiveBench	24.2%	#8 of 12	Official leaderboard
ScienceVision	0.36	#8 of 12	Official leaderboard
DataCenterBench Tier 1	15%	#6 of 9	Official leaderboard
Humanity's Sixth Sense	25.5%	#25 of 34	Official leaderboard
GDP.pdf	39.8%	#15 of 19	Official leaderboard
GlobeBench	14%	#16 of 19	Official leaderboard
Code2Video Arena	1,433	#11 of 13	Official leaderboard
Agentic Vision Bench (Moonshot)	78.3%	#1 of 1	Lab model card
KWV Bench	64.7%	#1 of 1	Lab model card

Safety and security

Benchmark	Score	Rank	Source
Molecular Biology Capabilities Test	58.4%	#12 of 90	Official leaderboard
Dictatorship Eval	90.3%	#3 of 20	Official leaderboard
CyberBench (Vals)	79%	#4 of 26	Official leaderboard
nocot-bench (NCRI)	96.8	#29 of 185	Official leaderboard
CyberBench v1.1 (Vals)	75.2%	#8 of 44	Official leaderboard
OncologyBench	24%	#6 of 30	Official leaderboard
Agentic BAIM-LLM Evaluation (ABLE)	78.4%	#4 of 18	Official leaderboard
ABC Bench Liquid-Handling Robot	99.3%	#8 of 34	Official leaderboard
ABC Bench Screening Evasion	69.1%	#4 of 17	Official leaderboard
World-Class Bio	55%	#18 of 74	Official leaderboard
Human Pathogen Capabilities Test	58.5%	#27 of 86	Official leaderboard
SecIT Bench	77%	#9 of 26	Official leaderboard
Human Pathogen Capabilities Test v2 (HPCT v2)	60.2%	#27 of 71	Official leaderboard
DistressBench	73.5%	#16 of 42	Official leaderboard
KORA	52.5%	#18 of 42	Official leaderboard
BioTIER Permit	99.5%	#32 of 69	Official leaderboard
HackBench	52.7	#4 of 8	Official leaderboard
DeepsecBench	17.4	#16 of 29	Official leaderboard
FORTRESS	26.6%	#41 of 73	Official leaderboard
SAFE-Teen Urgent safety	44%	#10 of 17	Official leaderboard
VariantBench	31.4%	#16 of 27	Official leaderboard
PCT (Polarized Contrastive Pairs)	47.6%	#28 of 46	Official leaderboard
dfbench	48%	#11 of 18	Official leaderboard
VCT-v2	41.8%	#16 of 25	Official leaderboard
TxBench-Antibody-Discovery	35.3%	#20 of 30	Official leaderboard
BioSecBench-Function	32.7%	#18 of 27	Official leaderboard
scBench-Long	14.2%	#18 of 27	Official leaderboard
CheatBench	70%	#7 of 10	Official leaderboard
ABC Bench Fragment Design	82.5%	#19 of 27	Official leaderboard
SAFE-Teen Relationship boundaries	63%	#12 of 17	Official leaderboard
BioTIER Refuse	27.3%	#49 of 69	Official leaderboard
BioSecBench-Refusal	8.30%	#16 of 22	Official leaderboard
SAFE-Teen Health advice	61%	#13 of 17	Official leaderboard
TxBench-Oligo-Discovery	31.3%	#23 of 30	Official leaderboard
MetagenomicsBench	27.3%	#20 of 26	Official leaderboard
VIALS 2	20.9%	#11 of 14	Official leaderboard
MiMo Cyber Bench	56.3%	#4 of 5	Official leaderboard
BioSecBench-Surveillance	38.6%	#22 of 27	Official leaderboard
TxBench-Preclinical-Pharmacology	39.7%	#25 of 30	Official leaderboard
CyberOffense	37.6%	#5 of 6	Official leaderboard
SpatialBench-Long	22.6%	#23 of 27	Official leaderboard
EpiBench	15.7%	#24 of 27	Official leaderboard
ExploitBench	32%	#19 of 21	Lab model card
Last Secure Code Benchmark	16.4%	#6 of 6	Official leaderboard
xdev-bench-hard	33.3%	#5 of 5	Official leaderboard
Kimi exploit development suite	38.9%	#1 of 1	Lab model card
UK AISI and CAISI simulated enterprise network range	17.0	#1 of 1	Lab model card

Other

Benchmark	Score	Rank	Source
SAGE (Vals)	52.8%	#5 of 81	Official leaderboard
Vals Multimodal Index	73.4%	#3 of 33	Official leaderboard
Xbench	58.1	#2 of 17	Official leaderboard
LMCA	52.7%	#24 of 147	Official leaderboard
Conceptual Reasoning Index	62.8	#30 of 151	Official leaderboard
CAIS Vision Capabilities Index	60.6	#15 of 53	Official leaderboard
CAIS Text Capabilities Index	40.5	#20 of 68	Official leaderboard
ACCoRD	70.5	#53 of 152	Official leaderboard
Artificial Analysis Intelligence Index v4.1	57.1	#3 of 8	Lab model card
SAGE v1.1 (Vals)	86.8%	#12 of 29	Official leaderboard
CompassRank	79.1	#6 of 14	Official leaderboard
Hermes Index	39.2	#8 of 17	Official leaderboard
LiveBench	65%	#18 of 31	Official leaderboard
CAIS Risk Index	53.6	#39 of 67	Official leaderboard

Its lab's numbers put it at 159.5, in line with its ECI

The Self-Reported Index fits one score to the 31 capability results Moonshot AI 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 156.7 to 162.3. How we compute it.

DeepSearchQA 95
MCPMark 94.5
MATH-Vision (MathVision) 94.3
GPQA Diamond 93.5
BrowseComp 91.2
Video-MME 90
Terminal-Bench 2.1 88.3
BabyVision 85.7
OSWorld-Verified 84.8
CharXiv 84.8
MCP Atlas 84.2
MMVU 82.1
MMMU-Pro 81.6
Toolathlon-Verified 76.5
AA-LCR (Artificial Analysis Long Context Reasoning) 74.7
DeepSWE 67.5
OfficeQA 63.3
SciCode 58.7
OSWorld 2.0 58.3
Humanity's Last Exam 56
Vals Finance Agent v2 54.4
JobBench 54.3
WorldVQA 51
SWE-Marathon 42
APEX-Agents 41
tau3-bench 33.4
τ³-bench Banking 33.4
AutomationBench 30.8
Agents' Last Exam 28.3
CritPt 23.4
ZeroBench 23

Its lab's numbers next to 17 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
ZeroBench	23% ZeroBench (pass@5) (without python)	41% Leaderboard, w/ tools; externally reported: Kimi K3 report; Python tools; 5 samplings.	−18 pts
AutomationBench	30.8% AutomationBench	22.7% Leaderboard	+8.1 pts
τ³-bench Banking	33.4% τ3-Banking, cited from Artificial Analysis (Table 2)	37.1% Leaderboard, max	−3.7 pts
Legal Research Bench	44.2% Legal Research Bench	46.1% Leaderboard	−1.9 pts
MCP Atlas	84.2% MCP-Atlas	82.3% Leaderboard, max	+1.9 pts
Vals Finance Agent v2	54.4% Finance Agent v2	53.1% Leaderboard	+1.3 pts
Code Arena	1,678 WebDev Arena (Elo), rank 1 of 99 (Table 5). Mapped to the current Arena WebDev board	1,658 Leaderboard, max	+20	Moonshot quotes Arena's July 23 rating, while the board reflects votes collected through November.
DeepSWE	67.5% DeepSWE (v1.1)	68.5% Leaderboard, max, mini-swe-agent	−1.0 pts
GPQA Diamond	93.5% GPQA Diamond	93.1% Epoch AI run, max	+0.4 pts
LMArena (Elo)	1,486 Text Arena (Elo), rank 8 of 200 (Table 5)	1,487 Leaderboard, max	−1	Moonshot quotes Arena's July 23 rating, while the board reports a later rating as votes accumulate.
CorpFin v2	71.6% CorpFin v2	71.6% Leaderboard	Same
Agents' Last Exam	28.3% Agents' Last Exam	28.3% Leaderboard, max, kimi_code	Same
BabyVision	85.7% BabyVision w/ python	85.7% Leaderboard, Python tools	Same
JobBench	54.3% JobBench	54.3% Leaderboard, OpenCode	Same
MLS-Bench	48.3% MLS-Bench-Lite	48.3% Leaderboard, max, Kimi-Code	Same
PerceptionBench (Moonshot)	58.5% PerceptionBench	58.5% Leaderboard	Same
Toolathlon-Verified	76.5% Toolathlon-Verified	76.5% Leaderboard, max	Same