Moonshot AI · China · Released 2025-07

Kimi K2

Moonshot's open agentic MoE (1T total / 32B active) tuned for tool use and long-horizon tasks.

Official site GitHub

Access

open-weights

Context

128,000 tokens

Modalities

text

Languages

40+

Benchmark scores

MMLU

84.1%

Massive Multitask Language Understanding

GPQA

68.4%

Graduate-level science QA (Diamond)

HumanEval

89.6%

Python code completion

SWE-bench

65.8%

Real GitHub issue resolution (Verified)

LiveCodeBench

N/A

Contamination-free coding

AIME

N/A

American Invitational Mathematics Exam

MMMU

N/A

Multimodal college-level reasoning

MATH

N/A

Competition-level math word problems

Scores as reported by the vendor or leading public leaderboards. "N/A" means the score has not been publicly disclosed for this metric.

Explain like I'm 5

An open model built to be an agent — it uses tools and plans multi-step tasks.

Key features

Agentic tuning
MoE
Tool use

Strengths

Strong SWE-bench for open weights
Great tool-use

Limitations

Text-only

Best for

Open-source agentsCoding assistants

Related models

OpenAI

GPT-5

OpenAI's flagship reasoning + multimodal model with a unified router that switches between fast and deep-thinking modes.

Anthropic

Claude Sonnet 4.5

Anthropic's best coding + agentic model, purpose-built for long-horizon computer-use and software engineering tasks.

Google DeepMind

Gemini 2.5 Pro

Google's most capable model with native 1M-token context and full multimodal I/O across text, images, audio and video.

xAI

Grok 4

xAI's frontier reasoning model with real-time X data grounding and a heavy multi-agent variant (Grok 4 Heavy).