Moonshot AI Open weights: moonshotai/Kimi-K3Reasoning modelCompare with another model

Kimi K3

Reasoning effort

Kimi K3 is a Moonshot AI open-weights reasoning model, released 16 Jul 2026. Its best configuration (max reasoning effort) ranks #42 of 775 on the BenchLeader Index at 64.9 ±3.3, in the upper half. The ± is the point: 113 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in composite (78) and lowest in agents & tools (60). At $6.00 per million tokens blended it is among the most expensive ranked models. Output speed of 42 tokens per second puts it in the slowest quarter, with a first answer in 51.7 s. It has been measured at 5 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 19 Oct 2026.

Blended price: $6.00/M; $3.00 in · $15.00 out
Output speed: 42 tok/s; measured by Artificial Analysis
First answer: 52 s; first token 1.65 s
Context: 1.0M
Full answer: 64 s; median, reasoning included
Cost per run: $2.00; one full Intelligence Index run
Released: 16 Jul 2026

How it scores by categoryDashed line = average model (50). One step of 15 = one standard deviation.

Overall index65
Reasoning67
Coding65
Agents & tools60
Maths64
Knowledge63
Human preference68
Multimodal64
Long context70
Composite78

Reasoning-effort configurations

The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.

Effort	Index	Rank	Speed	First answer	Chat reply cost	Categories
low	58.2	#159	39 tok/s	55 s	$0.0057	Agents & tools 71 · Coding 58 · Composite 63 · Knowledge 63 · Long context 65 · Maths 54 · Multimodal 62 · Reasoning 49
high	–	–	50 tok/s	1.65 s	$0.0057	Maths 64 · Reasoning 62
maxbest	64.9	#42	42 tok/s	52 s	$0.0057	Agents & tools 60 · Coding 65 · Composite 78 · Human preference 68 · Knowledge 63 · Long context 70 · Maths 64 · Multimodal 64 · Reasoning 67
thinking	–	–	50 tok/s	1.65 s	$0.0057	Maths 65
not stated	59.1	#135	50 tok/s	1.65 s	$0.0057	Agents & tools 55 · Coding 62 · Composite 45 · Human preference 71 · Knowledge 66 · Maths 72

How its index has moved

8 Sept 2026 to 20 Oct 2026

max
not stated
low

The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.

Benchmark results

See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.

Metadata

Benchmark	max	Source	Trend
AA Openness Indexnot in index	38.9#30	Artificial Analysis

Reasoning

Benchmark	low	high	max	not stated	Source
GPQA Diamond	84.8%#92	91.9%#27	93.1%#21	–	Epoch AI Benchmarking Hub
SimpleBench	–	–	60.7%#28	–	SimpleBench
Arena Hard Prompts	–	–	1516#11	–	Arena (LMArena)
LiveBench 2.0 Reasoningnot in indexeffort not stated	–	–	–	90.7%#10	LiveBench
LiveBench 1.0 Reasoningnot in indexeffort not stated	–	–	–	90.7%#9	LiveBench
GPQA Diamond (AA)not in index	84.2%#143	–	93.5%#14	–	Artificial Analysis
Humanity's Last Exam (AA)not in index	25.0%#182	–	46.9%#42	–	Artificial Analysis
GPQA Diamond (Vals)not in indexeffort not stated	–	–	–	92.9%#12	Vals AI
ARC-AGI-1	65.7%#141	86.7%#94	94.5%#37	–	ARC Prize
ARC-AGI-2	12.4%#146	55.0%#95	60.4%#81	–	ARC Prize
CritPt	3.1%#173	–	23.4%#49	–	Artificial Analysis
Chess Puzzlesnot in index	18.0%#98	37.0%#43	36.0%#48	–	Epoch AI Benchmarking Hub
Mystery Game Puzzlesnot in index	12.0%#102	–	26.0%#42	–	Epoch AI Benchmarking Hub
LMCAnot in index	–	–	52.7%#54	–	Epoch AI Benchmarking Hub
DTBenchnot in index	–	–	91.2%#73	–	Epoch AI Benchmarking Hub
ForecastBenchnot in index	–	–	64.6%#23	–	ForecastBench

Coding

Benchmark	low	max	not stated	Source
SciCode	52.7%#84	59.5%#13	58.7%#20	SciCode
WeirdML	–	82.6%#18	–	WeirdML
FrontierCodeeffort not stated	–	–	44.2%#16	Cognition
Arena Coding	–	1538#12	–	Arena (LMArena)
Arena WebDev	–	1655#15	–	Arena (LMArena)
LiveBench 2.0 Codingnot in indexeffort not stated	–	–	59.3%#43	LiveBench
LiveBench 1.0 Codingnot in indexeffort not stated	–	–	81.5%#14	LiveBench
SciCode (AA)not in index	52.7%#83	59.5%#14	–	Artificial Analysis
LiveCodeBench	–	87.2%#17	–	Vals AI
IOI	–	48.9%#25	–	Vals AI
SWE-bench (Vals)not in indexeffort not stated	–	–	93.4%#8	Vals AI
Code Migrationnot in index	–	16.1%#54	–	Vals AI
ProgramBenchnot in indexeffort not stated	–	–	2.0%#10	Vals AI
Vibe Code Bench 1-100not in index	–	18.2%#14	–	Vals AI
Vibe Code Bench v1.1not in index	–	85.0%#16	–	Vals AI
Surface Evolver Benchnot in index	–	93.0%#4	95.0%#1	Epoch AI Benchmarking Hub
DeepSWE v1.1not in index	–	68.5%#19	–	Epoch AI Benchmarking Hub
ALE-Benchnot in index	–	1524.5#22	–	Epoch AI Benchmarking Hub
FrontierSWEnot in index	–	25.9%#12	–	Epoch AI Benchmarking Hub
Terminal-Bench 4.0 (AA)not in index	12.6%#77	12.6%#77	–	Artificial Analysis
Terminal-Bench 2.1 (AA)not in index	82.4%#38	85.0%#26	–	Artificial Analysis
Startrise Frontendnot in indexeffort not stated	–	–	79.8%#4	Startrise
FrontierCode 1.1 (AA)not in index	–	43.6%#16	–	Artificial Analysis

Agents & tools

Benchmark	low	max	not stated	Source
APEX-Agentseffort not stated	–	–	47.8%#30	Mercor
Arena Agent	–	1.9#22	–	Arena (LMArena)
LiveBench Long Agenticnot in indexeffort not stated	–	–	42.8%#20	LiveBench
LiveBench Browsernot in indexeffort not stated	–	–	16.0%#39	LiveBench
LiveBench Agentic Codingnot in indexeffort not stated	–	–	62.2%#10	LiveBench
Terminal-Bench 2.1 (Vals)effort not stated	–	–	80.9%#9	Vals AI
MCP Atlas	–	82.3%#8	–	Scale AI SEAL
HiL-Bencheffort not stated	–	–	42.7%#14	Scale AI SEAL
τ³-bench Bankingnot in index	–	37.1%#11	–	τ²-bench
GDPval-AA v2.1not in index	31.3%#132	51.6%#46	–	Artificial Analysis
τ³-Banking (AA)	41.6%#25	46.0%#11	–	Artificial Analysis
ITBench SRE (AA)not in index	–	47.7%#9	–	Artificial Analysis
Analyst Agent (AA)not in index	–	38.8%#24	–	Artificial Analysis
APEX-Agents (AA)not in index	–	41.3%#3	–	Artificial Analysis
CyberBenchnot in index	–	75.2%#8	–	Vals AI
Finance Agent v2not in index	–	53.1%#34	–	Vals AI
Harvey's Legal Agent Benchmarknot in index	–	12.9%#10	–	Vals AI
Legal Research Benchnot in index	–	46.1%#16	–	Vals AI
Tax Agent Benchnot in index	–	33.4%#14	–	Vals AI
Terminal-Bench 4.0 (Vals)not in index	–	17.2%#31	–	Vals AI
Terminal-Bench Sciencenot in index	–	1.4%#35	–	Vals AI
Time Horizon Index: KSPnot in index	–	10.5%#9	–	Vals AI
PostTrainBenchnot in indexeffort not stated	–	–	32.0%#6	Epoch AI Benchmarking Hub
GBAEvalnot in indexeffort not stated	–	–	48.4%#9	Epoch AI Benchmarking Hub
Vending-Bench 2not in indexeffort not stated	–	–	5165.0#29	Andon Labs
AutomationBenchnot in index	49.3%#79	58.3%#43	–	Artificial Analysis
GDP.pdf (AA)not in index	18.6%#70	22.0%#50	–	Artificial Analysis
Harvey LAB-AA v1.1not in index	–	5.3%#8	–	Harvey
EnterpriseOps-Gymnot in index	–	45.3%#13	–	Artificial Analysis
AA-Briefcase v1.1not in index	–	1499#31	–	Artificial Analysis
CWE-Bench (AA)	–	53.3%#10	–	Artificial Analysis
DeepSec-Bench (AA)	–	12.4%#15	–	Artificial Analysis
CyberGym end-to-end (AA)	–	58.0%#7	–	Artificial Analysis

Maths

Benchmark	low	high	max	thinking	not stated	Source
FrontierMath Tiers 1–3	–	–	72.2%#25	–	–	Epoch AI Benchmarking Hub
FrontierMath Tier 4	–	–	39.0%#32	–	–	Epoch AI Benchmarking Hub
OTIS Mock AIME	68.9%#140	93.3%#58	97.2%#36	–	–	Epoch AI Benchmarking Hub
ProofBencheffort not stated	–	–	–	–	87.0%#10	Vals AI
LiveBench Mathematicsnot in indexeffort not stated	–	–	–	–	84.4%#55	LiveBench
AIME 2026possible contamination	–	–	–	96.7%#6	–	MathArena
HMMT February 2026possible contamination	–	–	–	97.0%#3	–	MathArena
MathArena Apexpossible contamination	–	–	–	65.6%#4	–	MathArena
Arena Maths	–	–	1501#16	–	–	Arena (LMArena)

Knowledge

Benchmark	low	max	not stated	Source
SimpleQA Verified	–	50.6%#26	–	Epoch AI Benchmarking Hub
LiveBench 2.0 Data Analysisnot in indexeffort not stated	–	–	78.7%#16	LiveBench
LiveBench Researchnot in indexeffort not stated	–	–	79.0%#20	LiveBench
LiveBench 1.0 Data Analysisnot in indexeffort not stated	–	–	78.7%#21	LiveBench
AA-Omniscience	3.9#113	19.7#69	–	Artificial Analysis
MMLU-Pro	–	88.0%#22	–	Vals AI
LegalBench	–	86.2%#16	–	Vals AI
CorpFineffort not stated	–	–	71.6%#3	Vals AI
TaxEvaleffort not stated	–	–	75.7%#12	Vals AI
BioMysteryBenchnot in index	–	72.6%#9	–	Vals AI
Excel Modeling Benchmarknot in index	–	66.7%#20	–	Vals AI
MedCodenot in index	–	49.4%#23	–	Vals AI
MedScribenot in index	–	88.0%#15	–	Vals AI
Public Benefits Benchnot in index	–	68.2%#11	–	Vals AI
GDP.pdfnot in index	–	19.0%#32	–	Epoch AI Benchmarking Hub
AA-Omniscience: accuracynot in index	45.8%#85	47.6%#74	–	Artificial Analysis
AA-Omniscience: non-hallucinationnot in index	22.9%#258	46.8%#136	–	Artificial Analysis

Instruction following

Benchmark	max	not stated	Source
LiveBench Languagenot in indexeffort not stated	–	85.5%#11	LiveBench
LiveBench Instruction Followingnot in indexeffort not stated	–	71.4%#24	LiveBench
Arena Instruction Followingnot in index	1488#15	–	Arena (LMArena)

Human preference

Benchmark	max	not stated	Source
Arena Text	1487#18	–	Arena (LMArena)
EQ-Bench 4effort not stated	–	1339#3	EQ-Bench
Arena Creative Writingnot in index	1461#27	–	Arena (LMArena)
Arena Multi-turnnot in index	1500#11	–	Arena (LMArena)
Arena Longer Queriesnot in index	1502#12	–	Arena (LMArena)

Multimodal

Benchmark	low	max	not stated	Source
LiveBench Designnot in indexeffort not stated	–	–	77.2%#22	LiveBench
MMMU-Pro	78.5%#72	80.5%#45	–	Artificial Analysis
MMMU-Pro (Vals)not in indexeffort not stated	–	–	88.2%#11	Vals AI
MortgageTaxnot in indexeffort not stated	–	–	66.3%#34	Vals AI
SAGEnot in index	–	52.8%#5	–	Vals AI
Blueprint-Bench 2not in indexeffort not stated	–	–	29.5%#19	Epoch AI Benchmarking Hub
AA-3Dnot in index	–	981#12	–	Artificial Analysis

Long context

Benchmark	low	max	Source	Trend
AA-LCR	79.3%#97	88.7%#1	Artificial Analysis
MLCRnot in index	–	38.3%#16	Artificial Analysis

Composite

Benchmark	low	max	not stated	Source
Epoch Capabilities Indexnot in indexeffort not stated	–	–	157.9#19	Epoch AI Benchmarking Hub
LiveBench 2.0effort not stated	–	–	63.4%#34	LiveBench
LiveBench 1.0not in indexeffort not stated	–	–	79.2%#15	LiveBench
AA Intelligence Index v4.3.2	30.1#122	43.6#45	–	Artificial Analysis
Vals Indexnot in index	–	53.0#17	–	Vals AI
Vals Multimodal Indexnot in indexeffort not stated	–	–	73.4%#3	Vals AI
AA Cyber Indexnot in index	–	41.3#7	–	Artificial Analysis

Safety & honesty

Benchmark	max	not stated	Source	Trend
FORTRESSnot in index	26.6%#29	27.3%#28	Scale AI SEAL

Where it wins

Benchmarks where this configuration ranks in the top five of every configuration measured.

#1AA-LCR88.7% · of 542

Where to run it

Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.

Provider	Speed	First token	Input $/M	Output $/M	Context	Quantisation
Parasail	97 tok/s	1.04 s	$4.50	$22.50	1M	fp4
Wafer	78 tok/s	0.81 s	$0.360	$11.90	1.0M	mxfp4
Wafer	77 tok/s	0.83 s	$0.238	$16.00	1.0M	mxfp4
Baseten	76 tok/s	2.90 s	$3.00	$15.00	1.0M	fp8
Fireworks	75 tok/s	1.90 s	$4.50	$22.50	1.0M	–
Fireworks	74 tok/s	0.38 s	$4.50	$22.50	1.0M	–
Amazon Bedrock	73 tok/s	2.63 s	$3.00	$15.00	1.0M	–
inference.net	71 tok/s	1.02 s	$2.24	$14.52	1.0M	fp4
inference.net	62 tok/s	1.17 s	$2.24	$14.52	1.0M	fp4
Sail Research	60 tok/s	1.65 s	$1.20	$5.50	1.0M	fp4
Sail Research	52 tok/s	2.46 s	$1.20	$5.50	1.0M	fp4
AkashML	52 tok/s	1.26 s	$1.20	$14.00	1.0M	fp4
Relace	51 tok/s	2.19 s	$0.300	$12.00	1.0M	fp4
Parasail	50 tok/s	1.47 s	$3.00	$15.00	1.0M	fp4
inference.net	50 tok/s	0.83 s	$0.239	$11.94	1.0M	fp4
inference.net	46 tok/s	1.21 s	$0.239	$11.94	1.0M	fp4
Phala	45 tok/s	2.60 s	$2.55	$12.75	1.0M	–
Morph	45 tok/s	2.48 s	$0.929	$14.90	1.0M	fp8
Together	40 tok/s	1.28 s	$3.00	$15.00	1.0M	–
DeepInfra	39 tok/s	0.97 s	$2.85	$14.25	1.0M	mxfp4
DigitalOcean	37 tok/s	2.06 s	$2.55	$12.95	1.0M	–
Decart	36 tok/s	5.51 s	$1.65	$8.25	1.0M	mxfp4
Alibaba Cloud Int.	35 tok/s	2.68 s	$3.45	$17.25	1.0M	–
Fireworks	32 tok/s	1.41 s	$3.00	$15.00	1.0M	–
Moonshot AI	26 tok/s	6.22 s	$3.00	$15.00	1.0M	mxfp4
Makora	25 tok/s	3.07 s	$0.650	$15.00	1.0M	fp4

Price history

Listed price per 1M tokens over time, as recorded by OpenRouter for the provider with the longest history.

input outputnow $1.00 in · $14.90 out

What a task costs

Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.300 per 1M. Reasoning tokens are not modelled.

Workload	Tokens in / out	Cost	With caching	Time
Chat reply	400 / 300	$0.0057	$0.0049	58.9 s
Summarize a 30-page report	12,000 / 600	$0.045	$0.021	1.1 min
Code edit	6,000 / 1,500	$0.041	$0.028	1.5 min
Agentic coding session	60,000 / 4,000	$0.240	$0.118	2.5 min
Structured extraction	2,000 / 200	$0.0090	$0.0050	56.5 s