Kimi K2.5
LatestMoonshot
•
Open Source•
1T# 35
Released
Jan 27, 2026
# 8
Context Length
256K
Benchmarks
# 54
Code RankedAGI
65.7%
# 32
SWEBench Pro
50.7%
# 10
SWEBench Multilingual
73.0%
# 20
SWEBench Verified
76.8%
# 30
Terminal Bench 2.0
50.8%
# 68
Agentic RankedAGI
54.7%
# 14
Browse Comp
74.9%
# 23
Code Arena
1427
# 9
Code Livebench
77.9%
# 13
AgenticCode LiveBench
48.3%
# 16
OSWorld Verified
63.3%
# 10
Svelte Bench
90.0%
# 9
LiveCodeBench v6
85.0%
# 11
Cyber Gym
41.3%
# 35
Reason RankedAGI
70.0%
# 32
HLE
30.1%
# 19
HLE w/ Tools
50.2%
# 25
GPQA Diamond
87.6%
# 4
AIME 2026
95.8%
# 8
AIME 2025 I & II
96.1%
# 19
Vending Bench 2
$1,198.46
# 7
NYT Connections
78.3%
# 7
MMLU Pro
87.1%
# 43
Halluc. Hughes
14.2%
# 10
Math RankedAGI
75.9%
# 51
RAGI RankedAGI
60.3%
# 2
DeepSearch QA
89.0%
# 28
GDPval AA
1288
# 18
Code DesignArena
1287
# 30
Toolathlon Pass@1
27.8%
# 8
Cursor Bench 3.1
31.9%
Pricing
# 25
Input Cost /M
$0.6
# 32
Output Cost /M
$3
# 12
Cached Cost /M
$0.1