DataLearner logo
KI

Kimi K2 Thinking

Reasoning modelKimi K2Kimi K2

Kimi K2 Thinking

Release date: 2025-11-06Updated: 2026-07-17Views: 2,956
Live demoGitHubHugging FaceCompare
Parameters
1T
Context length
256K
Multilingual
Supported
Reasoning ability
4/5

Kimi K2 Thinking is a reasoning model from Moonshot AI, released on 2025-11-06. It accepts text input and returns text output. The cataloged parameter count is 1T, with 32B active parameters per inference. The recorded context window is 256K. Cataloged capabilities include Reasoning model and Multilingual. The checkpoint is listed under the Modified Mit License, with the license link included in the model references. Use the linked references to confirm current access, licensing, and provider-specific limits.

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

Kimi K2 Thinking

Model basics

Reasoning traces
Supported
Thinking modes
Thinking Mode (Default)
Context length
256K tokens
Max output length
No data
Model type
Reasoning model
Modality (in / out)
Text → Text
Release date
2025-11-06
Model file size
1.09 TB
MoE architecture
Yes
Total params / Active params
1T / 32B
Knowledge cutoff
No data
Kimi K2 Thinking

Open source & experience

Code license
Modified MIT License
Weights license
Modified MIT License- Commercial use permitted
GitHub repo
github.com/MoonshotAI/Kimi-K2
Hugging Face
huggingface.co/moonshotai/Kimi-K2-Thinking
Live demo
kimi.com/en
Kimi K2 Thinking

Official resources

Paper
Paper
DataLearnerAI blog
DataLearnerAI blog
Kimi K2 Thinking

API details

API speed
3/5
💡Default unit: $/1M tokens. If vendors use other units, follow their published pricing.
Learn about pricing modes
Standard
TypeConditionInputOutput
Text-$0.600/ 1M$2.50/ 1M
Batch
TypeConditionInputOutput
Text-$0.300/ 1M$1.25/ 1M
Cache PricingPrompt Cache
TypeTTLWriteRead
Text-$0.600/ 1M$0.300/ 1M
Text-$0.600/ 1M
Cache = write
—
Text-—$0.300/ 1M
Cache = hit

“—” means the modality is not billed in that direction, or the vendor has not published a price for it.

Kimi K2 Thinking

Benchmark Results

Kimi K2 Thinking currently shows benchmark results led by AIME2025 (1 / 106, score 100), Terminal-Bench (4 / 35, score 47.10), HLE (27 / 185, score 51). 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
Parallel

General Knowledge

5 evaluations
Benchmark / mode
Score
Rank/total
MMLU Pro
Thinking Mode
84.60
34 / 133
LiveBench
Thinking Mode
61.59
63 / 115
HLE
Thinking Mode
23.90
120 / 185
HLE
Thinking ModeTools
44.90
47 / 185
HLE
HighTools
51
27 / 185

General Evaluation

1 evaluations
Benchmark / mode
Score
Rank/total
GPQA Diamond
Thinking Mode
84.50
106 / 270

Coding and Software Engineer

2 evaluations
Benchmark / mode
Score
Rank/total
LiveCodeBench
Thinking Mode
83.10
25 / 127
SWE-bench Verified
Thinking ModeTools
71.30
56 / 114

Math and Reasoning

4 evaluations
Benchmark / mode
Score
Rank/total
AIME2025
Thinking Mode
94.50
27 / 106
AIME2025
Thinking ModeTools
99.10
12 / 106
AIME2025
HighTools
100
1 / 106
FrontierMath - Tier 4
Thinking Mode
0
72 / 80

Writing and Creative Capabilities

1 evaluations
Benchmark / mode
Score
Rank/total
Creative Writing
Standard Mode
1627.80
31 / 99

AI Agent - Tool Usage

1 evaluations
Benchmark / mode
Score
Rank/total
Terminal-Bench
Thinking ModeTools
47.10
4 / 35

Common Sense Reasoning

1 evaluations
Benchmark / mode
Score
Rank/total
SimpleBench
Standard Mode
39.60
45 / 67

Agent Level Benchmark

1 evaluations
Benchmark / mode
Score
Rank/total
τ²-Bench - Telecom
Thinking ModeTools
93
18 / 35

Instruction Following

1 evaluations
Benchmark / mode
Score
Rank/total
IF Bench
Thinking Mode
68
21 / 34

AI Agent - Information Search

1 evaluations
Benchmark / mode
Score
Rank/total
BrowseComp
Thinking ModeTools
60.20
37 / 54

Claw-style Agent Evaluation

1 evaluations
Benchmark / mode
Score
Rank/total
Claw Bench
Thinking ModeTools
82.50
17 / 29
View benchmark analysisCompare with other models
Kimi K2 Thinking

Publisher

Kimi K2 Thinking

Model Overview

Kimi K2 Thinking is a reasoning model from Moonshot AI, released on 2025-11-06.

It accepts text input and produces text output. Its cataloged capabilities include Reasoning model and Multilingual. The cataloged parameter count is 1T, with 32B active parameters per inference. The recorded context window is 256K.

The checkpoint is listed under the Modified Mit License, with the license link included in the model references. The page records 8 API pricing rules from Fireworks AI; current provider pricing and conditions should be checked before deployment. The evaluation section contains 18 cataloged benchmark results with their recorded modes and scores. The page links 5 release, model-card, repository, or provider references for checking the underlying claims. Specifications, availability, and prices can change; undisclosed values are intentionally left unstated.

Kimi K2 Thinking

FAQ

What is Kimi K2 Thinking?

Kimi K2 Thinking is a reasoning model from Moonshot AI, released on 2025-11-06. It accepts text input and returns text output. The cataloged parameter count is 1T, with 32B active parameters per inference. The recorded context window is 256K. Cataloged capabilities include Reasoning model and Multilingual. The checkpoint is listed under the Modified Mit License, with the license link included in the model references. Use the linked references to confirm current access, licensing, and provider-specific limits.

What input and output modalities does Kimi K2 Thinking support?

The current model record lists text as input and text as output.

What are the main recorded specifications for Kimi K2 Thinking?

The cataloged parameter count is 1T, with 32B active parameters per inference. The recorded context window is 256K. Fields without a source-backed value remain undisclosed.

Does Kimi K2 Thinking have API pricing?

The page records 8 API pricing rules from Fireworks AI; current provider pricing and conditions should be checked before deployment.

Are benchmark results available for Kimi K2 Thinking?

The evaluation section contains 18 cataloged benchmark results with their recorded modes and scores. Compare only results that use the same benchmark version and evaluation mode.

Is Kimi K2 Thinking open source?

The checkpoint is listed under the Modified Mit License, with the license link included in the model references. Review the linked license text before commercial or derivative use.

DataLearner on WeChat

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

DataLearner WeChat QR code