NEWHow AI models draw the world map · Gecko Tests
Beta
Home/Models/Kimi K2 Thinking
moonshotai logo

Kimi K2 Thinking

by moonshotai · Released Nov 2025

Open Source
56.7
avg score
Rank #112
Compare
Better than 66% of all models
Context
262K tokens (~131 books)
Input $/1M
$0.60
Output $/1M
$2.50
Type
text
License
Open Source
Benchmarks
28 tested
Data as of
About

Kimi K2 Thinking is Moonshot AI’s most advanced open reasoning model to date, extending the K2 series into agentic, long-horizon reasoning. Built on the trillion-parameter Mixture-of-Experts (MoE) architecture introduced in...

Tested on 28 benchmarks · BenchGecko score 56.7. Top scores: OpenCompass · AIME2025 (94.1%), OpenCompass · IFEval (92.4%), OpenCompass · MMLU-Pro (84.3%).

Looking for similar performance at lower cost?
Qwen3 Next 80B A3B Thinking scores 56.8 (100% as good) at $0.15/1M input · 75% cheaper
Capabilities
coding
54.1
#78 globally
reasoning
47.8
#100 globally
math
55.9
#115 globally
knowledge
48.5
#148 globally
agentic
4.0
#85 globally
language
73.6
#74 globally
general
38.4
#89 globally
Benchmark Scores
Compare All
Tested on 28 benchmarks · Ranked across 7 categories
Score Distribution (all 327 models)
0255075100
▲ You are here
codingCompare coding →
OpenCompass · LiveCodeBenchV6

OpenCompass Live Code Bench v6. Fresh competitive programming problems to evaluate code generation without memorization.

77.1·
LiveBench · Coding

Regularly refreshed coding problems that avoid data contamination. New problems added monthly to prevent memorization.

67.4·
SWE-Bench Verified (Bash Only)

SWE-bench Verified solved using only bash commands, no specialized frameworks. Tests raw terminal-based problem solving.

63.4·
reasoningCompare reasoning →
LiveBench · Reasoning

Regularly refreshed reasoning problems testing logical deduction, spatial reasoning, and analytical thinking.

63.5·
LiveBench · Data Analysis

Fresh data analysis tasks testing ability to interpret tables, charts, and statistical data.

52.3·
SimpleBench

Deceptively simple questions that humans find easy but AI models often get wrong. Tests common sense and reasoning gaps.

27.5·
mathCompare math →
OpenCompass · AIME2025

OpenCompass evaluation on AIME 2025 problems. Tests mathematical reasoning on fresh competition problems.

94.1·
OTIS Mock AIME 2024-2025

Mock AIME (American Invitational Mathematics Exam) problems from OTIS. Tests mathematical competition performance.

83.0·
LiveBench · Mathematics

Regularly updated math problems that test numerical reasoning, algebra, calculus, and combinatorics.

81.1·
Quick compare:Alibaba Qwen logovs Qwen3 Next 80B A3B Thinkingwriter logovs Palmyra X5Anthropic logovs Claude Opus 4.5
Excellent (85+) Good (70-85) Average (50-70) Below (<50)
Recently Happened
Kimi K2 Thinking pricing increased 28%
Apr 9, 2026
Similar Models
Alibaba Qwen logo
Qwen3 Next 80B A3B Thinking
Alibaba Qwen
56.8$0.15/1M
writer logo
Palmyra X5
writer
57.0$0.60/1M
Anthropic logo
Claude Opus 4.5
Anthropic
56.4$5.00/1M
Links
Info
moonshotaiPricing explorerDevelopers · API
Research
Technical Report
Documentation
API DocsPlayground
Community
@moonshotai
Source Code
GitHubHugging Face
BenchGecko API
kimi-k2-thinking
Specifications
  • Typetext
  • Context262K tokens (~131 books)
  • ReleasedNov 2025
  • LicenseOpen Source
  • StatusActive
  • Cost / Message~$0.004
Available On
moonshotai logomoonshotai$0.60
Categories
codingreasoningmathknowledgeagenticlanguagegeneral
Learn More
context-windowtransformeropen weightstokens
Share & Export
Tweet
Related Models
Alibaba Qwen logoQwen3 Next 80B A3B Thinking
56.8
writer logoPalmyra X5
57.0
Anthropic logoClaude Opus 4.5
56.4
Alibaba Qwen logoQwen3.6 Max Preview
56.4
Kimi K2 Thinking is an open-source text AI model by moonshotai, released in November 2025. It has an average benchmark score of 56.7. Context window: 262K tokens.

Related Models

Qwen3 Next 80B A3B Thinking · Alibaba QwenPalmyra X5 · writerClaude Opus 4.5 · AnthropicQwen3.6 Max Preview · Alibaba QwenKimi K2 0711 · moonshotai

Benchmarks

OpenCompass · AIME2025OpenCompass · IFEvalOpenCompass · MMLU-ProOTIS Mock AIME 2024-2025OpenCompass · GPQA-Diamond

Related Pages

moonshotai · Providermoonshotai · EconomyAll ModelsCompare ModelsPricingDevelopers · APIContext Window · Glossary

Key facts · as of 2026-11-12

  • Kimi K2 Thinking by moonshotai. BenchGecko score 56.7, rank 112 of 327 scored models (normalized average of public benchmark scores).
  • List price $0.60 input · $2.50 output per 1M tokens (as of 2026-11-12).
  • Sold by 1 provider (as of 2026-11-12): Novita (bf16) $0.60 in / $2.50 out.
  • Gecko Tests: Who Are You B (Knows who made it) · Tokenizer Tax E (92% more tokens outside English).

How to cite · data as of 2026-11-12

Kimi K2 Thinking · benchmarks, pricing and providers. BenchGecko, data as of 2026-11-12. https://benchgecko.ai/model/kimi-k2-thinking

Credit "Source: BenchGecko" with a link. Prices per provider and Gecko Tests are BenchGecko data (CC BY 4.0); benchmark scores keep their original source, listed in the JSON. JSON · llms.txt · MCP