ModelBeat
Models/Kimi K2.5
All models
M

Moonshot: Kimi K2.5

moonshotai/kimi-k2.5

Kimi K2.5 is Moonshot AI's native multimodal model, delivering state-of-the-art visual coding capability and a self-directed agent swarm paradigm.

Context
262K
Input
$0.38 / 1M
Output
$2.02 / 1M
Released
Feb 2, 2026
Modalities
textimagetext

Benchmarks

Scores on standardized evaluations. Higher is better — and rank shows where Kimi K2.5 lands among all models tracked on Model Beat.

44.0
Intelligence Index
Epoch AI
44th percentile of tracked models
46.0
Coding Index
Epoch AI
46th percentile of tracked models
39.0
Agentic Index
Epoch AI
39th percentile of tracked models

Reasoning

7 evals
ARC-AGI65.3%

Abstract visual reasoning

ARC-AGI-211.8%

Harder abstract reasoning

GPQA Diamond87.6%

Graduate-level scientific reasoning

Humanity's Last Exam24.4%

Frontier of human expert knowledge

SimpleBench46.8%

Common-sense trick questions

SimpleQA Verified33.9%

Factual accuracy & hallucination

WeirdML45.6%

Novel ML problem-solving

Coding

2 evals
SWE-bench Verified73.8%

Real GitHub issue resolution

WebDev Arena1431

Human-rated web development

Math

3 evals
AIME 2024/202592.2%

Olympiad-qualifier math

FrontierMath27.9%

Research-level math problems

FrontierMath Tier 44.2%

Hardest research math

Agentic & Tools

2 evals
APEX14.4%

Multi-step agentic tasks

Terminal-Bench43.2%

Command-line agentic tasks

In the news

9

Model & benchmark data from Epoch AI (CC BY); pricing, specs & descriptions from OpenRouter.