0

Kimi K2 Thinking

Kimi K2 Thinking is an AI model from Kimi.

Publisher
Kimi
Family
Kimi
Context
262K
Input
Output
License
modified-mit
Released
6 Nov 2025

Cite

Notes

Only stored in your browser.

Attribution

Benchmark scores
frontiermathAALiveBenchSWE-benchVals AIEQ-Bench
Attribution policy →
Context
262K
Weights
Open
downloadable
Input
$0.60
/1M tok
Output
$2.50
/1M tok
Blended
$1.07
#389/ 499/1M · 3:1
Speed
0
#103/ 498tok/s
Cache read
$0.15
0.25x input

Price & speed via Artificial Analysis · ranked across tracked models · caching via OpenRouter

Reported on 22 evals across 9 domains - best 94.7% on AIME 2025: Problems from the American Invitational Mathematics Examination · #9 of 205

Reported eval scores

22
AIME 2025: Problems from the American Invitational Mathematics Examination94.7
LiveCodeBench85.3
SWE-bench Verified63.4
MMLU-Pro84.8
Humanity's Last Exam (HLE)23.8
LiveBench65.5
τ²-bench (Tau²-bench)93
IFBench68.1
Terminal-Bench (Hard)31.1
LiveBench - Language66.5
LiveBench - Coding67.4
LiveBench - Math81.1
LiveBench - Instruction Following62
LiveBench - Data Analysis52.3
LiveBench - Reasoning63.5
GPQA Diamond83.8
SciCode42.4
EQ-Bench Creative Writing v31630.9
EQ-Bench Longform Creative Writing69
FrontierMath21.4