Moonshot AI

Kimi Code

#28of 28 · general

Kimi K2.7 Code

1.0Tparams · MoE

262kcontext

$0.95 / $4per 1M in / out

2026-06-12released

weights ↗repo ↗

Coding11#27

Agents11#27

Cyber—

Engineering—

Math2#26

Health—

Legal—

Professional3#28

Writing—

Instruct4#25

Hallucination11#25

Multimodal—

History

Moonshot AI
older → newer

$ / M input tokens

Epoch Capabilities Index against input price
Model	$ / M input	ECI points	Released
Kimi K2	—	140.0	2025-07-12
Kimi K2 Thinking	—	145.8	2025-11-06
Kimi K2.5	—	148.1	2026-01-27
Kimi K2.6	—	150.9	2026-04-20
Kimi K2.7 Code	0.6712	150.1	2026-06-12

Usage

1.9TtokensAug 2 – Oct 6▼ 7% 7d

Geography

United States10.3%

China6.8%

Brazil5.8%

Spain5.0%

Germany4.8%

India4.0%

Colombia3.7%

Argentina3.1%

France2.6%

Indonesia2.6%

Measured

≈ imputed · amber dot = lab-self-reported among the sources

GPQA Diamond87.9%#22

HLE35.0%#25

LiveBench70.2%#20

SWE-bench V78.2%#8

LiveCodeBench82.0%#8

SciCode47.8%#24

CritPt10.0%#25

Terminal-Bench67.0%#14

APEX Agents37.6%#19

GDPval26.9%#24

AA Index25.8#26

Epoch Capabilities Index150.1#19

Neighbors

MiniMax→ more factual · cheaper Mistral→ stronger math · stronger coding Qwen Lite→ broader · cheaper MiMo Flash→ stronger agent · cheaper GPT Luna→ stronger math · cheaper