Kimi K2 Thinking
Developed by Moonshot AI, Kimi K 2 Thinking is an open-source chat model that excels at multi-step reasoning and tool orchestration, enabling autonomous workflows such as research, coding, and writing. Its notable technical trait is the use of a Mixture-of-Experts (MoE) architecture with a large context window of 262,144 tokens.
Input
Output
Context
262K
Max Output
262K
Parameters
1026.4B
Technical Specifications
Model TypeChat
Context Window262,144 tokens
Max Output Tokens262,144 tokens
Parameters1026.4B
Release DateNov 4, 2025
Training CutoffNot available
Licenseother
Open SourceYes
Input Modalities
Text
Output Modalities
Text
Capabilities
Benchmarks
Artificial Analysis
23.8%
HLE
70.3%
AA-LCR
83.8%
GPQA
93.0%
Tau2-Bench
94.7%
AIME 2025
68.1%
IFBench
42.4%
SciCode
84.8%
MMLU Pro
94.7
Math
85.3%
LiveCodeBench
33.5
Intelligence
31.1%
Terminal-Bench Hard
More Models by Moonshot AI
Kimi Linear 48B A3B Base1000K context · 49.1B · Open weightsKimi K2.6262K context · 1026.9B · Open weightsMoonlight 16B A3B Instruct8K context · 16B · Open weightsKimi K2 Instruct131K context · 1026.4B · Open weightsKimi K2.7 Code262K context · 1026.9B · Open weightsKimi Dev 72B131K context · 72.7B · Open weights
Related Tools
Data sourced from official provider APIs and documentation
Last updated: Aug 28, 2026
Your AI stack shouldn't stand still.
Every month new models become cheaper, faster, and more capable. Inferbase ensures your application automatically benefits without changing a single API call.