DataLearner logo
KI

Kimi K3

Reasoning modelCoding modelKimi K3

Kimi K3

Release date: 2026-07-16Updated: 2026-07-17 08:50:06.750831
Live demoGitHubHugging FaceCompare
Parameters
2.8T
Context length
1M
Chinese support
Not supported
Reasoning ability

Kimi K3 是 Moonshot AI 的 2.8T 参数开放式 MoE 旗舰模型，拥有 1M 上下文、原生视觉和 16/896 专家激活；官方公布了 35 条 max 档评测，并计划最迟于 2026-07-27 发布完整权重。API 价格为缓存输入 ¥2/$0.30、非缓存输入 ¥20/$3、输出 ¥100/$15（每百万 tokens）。

Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology

Kimi K3

Model basics

Reasoning traces
Supported
Thinking modes
Thinking Level · Max (Default)
Context length
1M tokens
Max output length
1M tokens
Model type
Reasoning model
Modality (in / out)
Text, Image, Video → Text
Release date
2026-07-16
Model file size
No data
MoE architecture
Yes
Total params / Active params
2.8T / N/A
Knowledge cutoff
No data
Kimi K3

Open source & experience

Code license
不开源
Weights license
不开源
GitHub repo
GitHub link unavailable
Hugging Face
Hugging Face link unavailable
Live demo
https://www.kimi.com/
Kimi K3

Official resources

Kimi K3

API details

API speed
No data
💡Default unit: $/1M tokens. If vendors use other units, follow their published pricing.
Learn about pricing modes
Standard
TypeConditionInputOutput
Text-¥20.00/ 1M¥100.00/ 1M
international
TypeConditionInputOutput
Text-$3.00/ 1M$15.00/ 1M
Cache PricingPrompt Cache
TypeTTLWriteRead
Text--$0.300/ 1M
Kimi K3

Benchmark Results

Kimi K3 currently shows benchmark results led by BrowseComp (1 / 51, score 91.20), GPQA Diamond (8 / 182, score 93.50), HLE (10 / 168, score 56). This page also consolidates core specs, context limits, and API pricing so you can evaluate the model from benchmark results and deployment constraints together.

Thinking
Tool usage
Internet

General Knowledge

3 evaluations
Benchmark / mode
Score
Rank/total
GPQA Diamond
Max
93.50
8 / 182
HLE
Max
43.50
44 / 168
HLE
MaxTools
56
10 / 168

AI Agent - Information Search

2 evaluations
Benchmark / mode
Score
Rank/total
DeepSearchQA
MaxToolsInternet
95
1 / 1
BrowseComp
MaxToolsInternet
91.20
1 / 51

AI Agent - Tool Usage

4 evaluations
Benchmark / mode
Score
Rank/total
TerminalBench 2.1
MaxTools
88.30
2 / 23
MCP-Atlas
MaxTools
84.20
2 / 27
Toolathlon-Verified
MaxTools
73.20
1 / 1
Automation Bench
MaxTools
30.80
1 / 1

Coding and Software Engineer

7 evaluations
Benchmark / mode
Score
Rank/total
FrontierSWE
MaxTools
81.20
1 / 1
Program Bench
MaxTools
77.80
1 / 1
Kimi Code Bench 2.0
MaxTools
72.90
1 / 1
DeepSWE
MaxTools
67.50
4 / 17
MLS Bench
MaxTools
48.30
1 / 1
SWE-Marathon
MaxTools
42
1 / 2
PostTrain Bench
MaxTools
36.60
1 / 1

Productivity Knowledge

5 evaluations
Benchmark / mode
Score
Rank/total
GDPval-AA v2
MaxTools
1668
1 / 1
AA-Briefcase
MaxTools
1548
1 / 1
DECK-Bench
MaxTools
73.50
1 / 1
Office QA Pro
MaxTools
63.30
1 / 1
SpreadsheetBench 2
MaxTools
34.80
1 / 1

Agent Level Benchmark

2 evaluations
Benchmark / mode
Score
Rank/total
Job Bench
MaxTools
52.90
1 / 1
APEX-Agents
MaxTools
37.60
1 / 1

Multimodal Understanding

12 evaluations
Benchmark / mode
Score
Rank/total
MathVision
Max
94.30
2 / 2
MathVision
MaxTools
97.80
1 / 2
CharXiv RQ
Max
84.80
2 / 2
CharXiv RQ
MaxTools
91.30
1 / 2
OmniDocBench
Max
91.10
1 / 1
BabyVision
MaxTools
85.70
1 / 1
MMMU-Pro
Max
81.60
2 / 2
MMMU-Pro
MaxTools
83.40
1 / 2
PerceptionBench
Max
58.50
1 / 1
WorldVQA ForceAnswer
Max
51
1 / 1
ZeroBench Main
Max
23
2 / 2
ZeroBench Main
MaxTools
41
1 / 2
View benchmark analysisCompare with other models

Compare with other models

No curated comparisons for this model yet.

Want a custom combination? Open the compare tool

Kimi K3

Publisher

Kimi K3

Model Overview

Kimi K3 是 Moonshot AI 的 2.8T 参数开放式 MoE 旗舰模型，拥有 1M 上下文、原生视觉和 16/896 专家激活；官方公布了 35 条 max 档评测，并计划最迟于 2026-07-27 发布完整权重。API 价格为缓存输入 ¥2/$0.30、非缓存输入 ¥20/$3、输出 ¥100/$15（每百万 tokens）。

DataLearner on WeChat

Follow DataLearner on WeChat for AI model updates and research notes.

DataLearner WeChat QR code