Kimi K2.5 is Moonshot AI's native multimodal model released in January 2026, delivering state-of-the-art visual coding capability and a self-directed agent swarm paradigm. Built on Kimi K2 with continued pretraining over approximately 15 trillion mixed visual and text tokens, it generates code from visual specifications — turning UI designs and video workflows into working implementations. Its agent swarm technology can self-direct up to 100 parallel sub-agents, each independently using tools to search, generate, analyze, and organize information, reducing execution time by up to 4.5× for complex research and writing tasks.
API|VisionReasoning|Open ModelModified MIT
Knowledge Cutoff
2026-02-02
Input → Output Format
Context Memory
262KIN66KOUT
AI Performance Evaluation
Arena Overall Score
1450
±4As of 2026-05-01
Overall Rank
No.40
26,123 Votes
Arena by Ability
Hard Prompts
1471±5No.42
Expert Knowledge
1486±14No.28
Instruction Following
1438±7No.40
Conversation Memory
1452±9No.46
Creative
1415±10No.48
Coding
1507±8No.30
Math
1474±14No.17
Arena by Occupation
Creative Writing
1424±8No.49
Social Sciences
1469±9No.37
Media
1421±9No.42
Business
1436±9No.57
Healthcare
1465±14No.53
Legal
1444±13No.62
Software
1492±6No.32
Mathematics
1480±16No.15
Source:Arena Intelligence
Overall
AA Intelligence Index
47%↑8%
LiveBench
69%↑8%
Reasoning & Math
GPQA Diamond
88%↑6%
HLE
29%↑12%
LB Reasoning
76%↑7%
LB Math
85%↑11%
LB Data
61%↑8%
Coding
AA Coding Index
40%↑3%
LB Coding
78%↑5%
LB Agentic
48%↑3%
TAU2
96%↑15%
TerminalBench
35%↑1%
SciCode
49%↑7%
Language & Instructions
IFBench
70%↑7%
AA-LCR
65%↑3%
Hallucination (HHEM)
14%↑4%
Factual (HHEM)
86%↓4%
LB Language
78%↑5%
LB IF
57%↑6%
Output Speed
Standard Mode
48tok/s↓30
First Output 1.26s
Reasoning Mode
46tok/s↓41
First Output 66.42s
