MoonshotAI: Kimi K2.7 Code

Moonshot

specialist_code

Capabilities

VisionToolsStructured OutputExtended ThinkingLong ContextCode

Strengths

Specialist code fitness (0.79) with extended thinking
262K context window for long-file and multi-file tasks
Open weights enable self-hosting and data sovereignty
Multimodal with vision support for diagram-based coding

Weaknesses

Very slow inference (0.04 speed_score; ~46 tok/s)
High output pricing ($3.5/M) for reasoning overhead
Limited non-code task specialisation
Privacy trade-off: closed training data despite open weights

Pricing

Input / 1M tokens

$0.74

Output / 1M tokens

$3.5

Context window

262.144k

Transparency

Open weights

10.0 / 10

Open training data

0.0 / 10

Open methodology

6.5 / 10

Licence openness

9.0 / 10

Provider disclosure

5.5 / 10

FMTI company score

N/A

Composite:6.0 / 10

Moonshot ships open weights for Kimi K2.7, enabling self-hosting and full data control. Training methodology partially disclosed; training data provenance not fully transparent. Open licence supports community use and auditing.

Sustainability

Inference energy

N/A

Training footprint

N/A

Provider infrastructure

5.0 / 10

Composite:5.0 / 10

No public green infrastructure commitments or energy efficiency data disclosed. Large model size (262K context) suggests higher inference costs; sustainability claims unavailable.

Task Fitness

question answering

69%

code

79%

mathematics

50%

business communication

50%

vision

68%

analysis

75%

extraction

72%

long-form generation

74%

research

50%

reasoning

77%

summarisation

72%

translation

68%

conversation

65%