What is the IQ of kimi-k3?

kimi-k3 has an estimated IQ of 127, which ranks #20 out of 164 public models on the AI IQ leaderboard. The score combines seven equally weighted reasoning dimensions using measured benchmark results and conservative estimates for missing coverage.

What is kimi-k3 best at?

Among the dimensions with published benchmark results, kimi-k3 ranks highest in Computer Use (IQ 130, #12 of 164) and Academic Reasoning (IQ 132, #20 of 181). Its lowest-ranked measured dimension is Reliability (IQ 121, #35 of 164).

What is the EQ of kimi-k3?

kimi-k3 has an estimated EQ of 133. EQ measures emotional reasoning — how well a model reads and responds to emotional context — and is scored separately from IQ.

Which benchmarks has kimi-k3 been tested on?

kimi-k3 has published results on 31 of the 39 benchmarks behind AI IQ's Composite IQ, including AA Long Context Reasoning v1.1, BrowseComp, SciCode, GPQA Diamond and BullshitBench v2.

How much does kimi-k3 cost?

kimi-k3 costs $3.00 per million input tokens and $15.00 per million output tokens. Adjusted for how many tokens it actually consumes on real tasks, its effective cost works out to $19.85 per million input-plus-output tokens.

Is kimi-k3 open source?

Yes — kimi-k3 is an open-weights model: its trained weights are published, so it can be downloaded and self-hosted.

Who developed kimi-k3?

kimi-k3 was developed by Kimi, based in China.

When was kimi-k3 released?

kimi-k3 was publicly released on July 16, 2026.

How large is the context window of kimi-k3?

kimi-k3 supports a context window of 1,048,576 tokens — the amount of text it can consider in a single request.

In independent measurements, kimi-k3 generates about 43 tokens per second, with a typical response time of about 61 seconds.

KimiReleased Jul 16, 2026

Open weights

kimi-k3

Strongest in Computer Use (#12 of 164) and Academic Reasoning (#20 of 181); weakest in Reliability (#35 of 164). All seven dimensions are measured.

Composite IQ ?

127

#20 of 164 ranked models · ahead of 88%

Kimi's top-ranked model

Effective cost

$19.85/1M

$3 in · $15 out

Speed

43tok/s

61s typical response

Context

1Mtokens

1,048,576 tokens in

Evidence

7of 7 dims

31 of 39 benchmarks

Compare with gpt-6.1-solAPI pricing

IQ profile

Seven dimensions, averaged equally into Composite IQ

Compare shape with

Measured95% range

#1 in dimensionTop-10 median

IQ dimensions

Ranks are among models scored in each dimension. Ranges are model-based estimates, not calibrated confidence intervals.

DimensionIQPosition · 70–160RankEvidencevs Top 10

Abstract ReasoningAbstract

106measured

#32 / 181

4/5

-20

Mathematical ReasoningMath

136measured

#19 / 164

4/5

-11

Academic ReasoningAcademic

132measured

#20 / 181

5/6

-3.5

Programmatic ReasoningCoding

135measured

#24 / 165

3/6

-14

Computer UseComputer Use

130measured

#12 / 164

4/5

-10.5

ReliabilityReliability

121measured

#35 / 164

5/6

-10.5

Knowledge WorkKnowledge Work

127measured

#19 / 164

6/6

-5

kimi-k395% range (model-based, uncalibrated)Top-10 median#1 in dimension100 = average human

Benchmark results

31 of 39 Composite IQ benchmarks have a published result for kimi-k3. Benchmarks without a result stay listed so gaps are visible.

Abstract Reasoning IQ 106 · measuredBest: #15 of 61 on ARC-AGI-14/5

BenchmarkScorePercentileRankLeader · source

ARC-AGI-1

ARC Prize verified leaderboard · Oct 13

94.5%

#15 / 61

Leader fable-5 · 98.5%
ARC Prize verified leaderboard · Oct 13

Chess Puzzles

Epoch AI vEpoch 100-position leaderboard (task implementation versions recorded per run) · Oct 13

39.0%

#21 / 76

Leader gpt-6-astra · 72.0%
Epoch AI vEpoch 100-position leaderboard (task implementation versions recorded per run) · Oct 13

ARC-AGI-2

ARC Prize verified leaderboard · Oct 13

60.4%

#28 / 61

Leader gpt-6-astra · 95.0%
ARC Prize verified leaderboard · Oct 13

Mystery Game Puzzles

Epoch AI vEpoch 100-position leaderboard (task implementation versions recorded per run) · Oct 13

26.0%

#32 / 55

Leader gpt-6-astra · 84.0%
Epoch AI vEpoch 100-position leaderboard (task implementation versions recorded per run) · Oct 13

Not yet reported

ARC-AGI-3

Mathematical Reasoning IQ 136 · measuredBest: #12 of 57 on MathArena4/5

BenchmarkScorePercentileRankLeader · source

MathArena

MathArena Models · Oct 13

50.5%

#12 / 57

Leader gpt-6.1-sol · 90.1%
MathArena Models · Oct 13

FrontierMath Tier 1-3

Epoch AI FrontierMath Tiers 1-3 v2 · Oct 13

72.2%

#18 / 78

Leader gpt-6-astra · 93.7%
Epoch AI FrontierMath Tiers 1-3 v2 · Oct 13

ProofBench 1.1

ProofBench 1.1 · Oct 13

87.0%

#11 / 46

Leader fable-5.1 · 100%
ProofBench 1.1 · Oct 13

FrontierMath Tier 4

Epoch AI FrontierMath Tier 4 v2 · Oct 13

39.0%

#18 / 66

Leader gpt-6.1-sol · 100%
Epoch AI FrontierMath Tier 4 v2 · Oct 13

Not yet reported

AIME

Academic Reasoning IQ 132 · measuredBest: #11 of 157 on SciCode5/6

BenchmarkScorePercentileRankLeader · source

SciCode

Artificial Analysis v4.3.2 · Oct 13

59.5%

#11 / 157

Leader opus-5.5 · 66.9%
Artificial Analysis v4.3.2 · Oct 13

GPQA Diamond

Artificial Analysis v4.3.2 · Oct 13

93.5%

#13 / 145

Leader gpt-6-astra · 96.1%
Artificial Analysis v4.3.2 · Oct 13

CritPt

Artificial Analysis v4.3.2 · Oct 13

23.4%

#18 / 149

Leader Muse Wonder Full · 32.9%
Artificial Analysis v4.3.2 · Oct 13

Humanity's Last Exam

Artificial Analysis v4.3.2 · Oct 13

46.9%

#20 / 157

Leader opus-5.5 · 61.4%
Artificial Analysis v4.3.2 · Oct 13

MMMU-Pro

Artificial Analysis v4.3.2 · Oct 13

80.5%

#18 / 84

Leader fable-5.1 · 90.6%
Artificial Analysis v4.3.2 · Oct 13

Not yet reported

MMLU-Pro

Programmatic Reasoning IQ 135 · measuredBest: #15 of 61 on ProgramBench3/6

BenchmarkScorePercentileRankLeader · source

ProgramBench

Vals.ai · Oct 13

17.5%

#15 / 61

Leader opus-5.5 · 65.0%
Vals.ai · Oct 13

Terminal-Bench 4.0

Artificial Analysis v4.3.2 · Oct 13

12.6%

#33 / 95

Leader sonnet-5.5 · 63.6%
Artificial Analysis v4.3.2 · Oct 13

IOI V2

Vals.ai v2 · Oct 13

48.9%

#25 / 41

Leader Gemini 4 Argon · 100%
Vals.ai v2 · Oct 13

Not yet reported

LiveCodeBenchSWE-rebenchMirrorCode

Computer Use IQ 130 · measuredBest: #2 of 39 on BrowseComp4/5

BenchmarkScorePercentileRankLeader · source

BrowseComp

LLM Stats BrowseComp · Oct 13

91.2%

#2 / 39

Leader gpt-6-astra · 91.5%
LLM Stats BrowseComp · Oct 13

MCP Atlas

Scale Labs MCP Atlas · Oct 13

82.3%

#8 / 32

Leader muse-spark-1.1 · 88.1%
Scale Labs MCP Atlas · Oct 13

Agents' Last Exam

Agents' Last Exam · Oct 13

28.3%

#8 / 29

Leader opus-5.5 · 38.2%
Agents' Last Exam · Oct 13

Arena.ai Agent Arena

Arena.ai Agent Arena · Oct 13

3.0

#22 / 59

Leader opus-5.5 · 17.9
Arena.ai Agent Arena · Oct 13

Not yet reported

OSWorld-Verified

Reliability IQ 121 · measuredBest: #1 of 144 on AA Long Context Reasoning v1.15/6

BenchmarkScorePercentileRankLeader · source

AA Long Context Reasoning v1.1

Artificial Analysis v4.3.2 · Oct 13

88.7%

#1 / 144

Leader on this benchmark
Artificial Analysis v4.3.2 · Oct 13

BullshitBench v2

BullshitBench v2 · Oct 13

74.0%

#8 / 81

Leader opus-4.8 · 94.0%
BullshitBench v2 · Oct 13

AA Omniscience

Artificial Analysis v4.3.2 · Oct 13

19.7

#27 / 148

Leader opus-5.5 · 46.4
Artificial Analysis v4.3.2 · Oct 13

SimpleQA Verified

Epoch AI SimpleQA Verified · Oct 13

50.6%

#27 / 75

Leader gpt-6-astra · 75.6%
Epoch AI SimpleQA Verified · Oct 13

IFBench

Artificial Analysis model leaderboard · Aug 19

23.0%

#95 / 95

Leader minimax-m3 · 82.9%
Artificial Analysis model leaderboard · Aug 19

Not yet reported

FACTS Grounding

Knowledge Work IQ 127 · measuredBest: #17 of 96 on GDP.pdf6/6

BenchmarkScorePercentileRankLeader · source

GDP.pdf

Artificial Analysis v4.3.2 · Oct 13

22.0%

#17 / 96

Leader Muse Wonder Full · 32.6%
Artificial Analysis v4.3.2 · Oct 13

AA-Briefcase v1.1 Elo

Artificial Analysis v4.3.2 · Oct 13

1,501

#18 / 100

Leader sonnet-5.5 · 1,823
Artificial Analysis v4.3.2 · Oct 13

AutomationBench-AA

Artificial Analysis v4.3.2 · Oct 13

58.3%

#21 / 97

Leader Gemini 4 Argon · 77.5%
Artificial Analysis v4.3.2 · Oct 13

GDPval-AA v2.1 normalized score

Artificial Analysis v4.3.2 · Oct 13

51.8%

#26 / 114

Leader opus-5.5 · 68.3%
Artificial Analysis v4.3.2 · Oct 13

EnterpriseOps-Gym-AA

Artificial Analysis v1117-task-strict-2026-09-29 · Oct 13

45.3%

#8 / 19

Leader fable-5 · 51.1%
Artificial Analysis v1117-task-strict-2026-09-29 · Oct 13

AA-AnalystAgent

Artificial Analysis vpass^5-2026-09-29 · Oct 13

38.8%

#15 / 23

Leader gemini-3.7-flash · 60.0%
Artificial Analysis vpass^5-2026-09-29 · Oct 13

Release & availability

Release details on record for this model.

Released: Jul 16, 2026
Developer: Kimi
Weights: Open weightsDownloadable and self-hostable
Context: 1,048,576 tokens
API price: $3 in · $15 outper 1M tokens · full pricing

About this score

How much of kimi-k3's IQ is measured, and how the rest is estimated.

31/ 39benchmark results

7/ 7dimensions measured

0estimated

4/5Abstract

4/5Math

5/6Academic

3/6Coding

4/5Computer Use

5/6Reliability

6/6Knowledge Work

Bar height = share of benchmarks with resultsEstimatedNo score

Measured results are shown as published

Every score in the benchmark tables comes from its source unchanged. Composite IQ averages all seven dimensions equally.

How gaps are handled

Every dimension has at least one published result.

A missing result is not a zero

Benchmark versions and evaluation harnesses can differ between sources, so small gaps between models may not be meaningful.

Read the scoring methodology →

Compare kimi-k3

Side-by-side pages for the models it is most often weighed against.

#1 overall

gpt-6.1-sol

IQ142 15 higherEffective cost$6.20 3.2× cheaper

Closest IQ

gemini-3.7-flash

IQ127 same IQEffective cost$5.95 3.3× cheaper

Most similar profile

sonnet-5

IQ125 2 lowerEffective cost$37.90 1.9× pricier

Similar price

Qwen3.8 Max 0902

IQ126 1 lowerEffective cost$20.14 about the same

#2 overall

opus-5.5

IQ142 15 higherEffective cost$35.08 1.8× pricier

#3 overall

gpt-6-astra

IQ140 13 higherEffective cost$33.53 1.7× pricier

kimi-k3Comparison model

Compare with any modelBrowse all comparisons →

Frequently asked questions

Quick answers about kimi-k3, drawn from the same data as the rest of this page.

How AI IQ scores models →Get this data from the API →

What is the IQ of kimi-k3?: kimi-k3 has an estimated IQ of 127, which ranks #20 out of 164 public models on the AI IQ leaderboard. The score combines seven equally weighted reasoning dimensions using measured benchmark results and conservative estimates for missing coverage.
What is kimi-k3 best at?: Among the dimensions with published benchmark results, kimi-k3 ranks highest in Computer Use (IQ 130, #12 of 164) and Academic Reasoning (IQ 132, #20 of 181). Its lowest-ranked measured dimension is Reliability (IQ 121, #35 of 164).
What is the EQ of kimi-k3?: kimi-k3 has an estimated EQ of 133. EQ measures emotional reasoning — how well a model reads and responds to emotional context — and is scored separately from IQ.
Which benchmarks has kimi-k3 been tested on?: kimi-k3 has published results on 31 of the 39 benchmarks behind AI IQ's Composite IQ, including AA Long Context Reasoning v1.1, BrowseComp, SciCode, GPQA Diamond and BullshitBench v2.
How much does kimi-k3 cost?: kimi-k3 costs $3.00 per million input tokens and $15.00 per million output tokens. Adjusted for how many tokens it actually consumes on real tasks, its effective cost works out to $19.85 per million input-plus-output tokens.
Is kimi-k3 open source?: Yes — kimi-k3 is an open-weights model: its trained weights are published, so it can be downloaded and self-hosted.
Who developed kimi-k3?: kimi-k3 was developed by Kimi, based in China.
When was kimi-k3 released?: kimi-k3 was publicly released on July 16, 2026.
How large is the context window of kimi-k3?: kimi-k3 supports a context window of 1,048,576 tokens — the amount of text it can consider in a single request.
How fast is kimi-k3?: In independent measurements, kimi-k3 generates about 43 tokens per second, with a typical response time of about 61 seconds.