LANGUAGE MODEL Moonshot AI Last updated:

Kimi K2

Moonshot's Open MoE Reasoning

Moonshot AI's first open-weight foundation model, released July 2025 — a strategic pivot from the closed-only Kimi consumer product positioning. Trillion-parameter MoE, 32B active, competitive with DeepSeek V3 on most benchmarks and with Claude Sonnet 4 on agentic coding. Brought Moonshot back into the open-frontier conversation after a difficult 2024 of consumer-product user growth challenges.

Intelligence
Good
Speed
Slow
41 tok/s output
Cost
Moderate
$0.50 in / $2.85 out
Context
200K
Up to 200,000 tokens
How are Intelligence, Speed & Cost bucketed?
Intelligence and Speed buckets are percentile ranks on Artificial Analysis. Cost buckets are fixed dollar thresholds keyed off output-token price ($/M out).
Intelligence
  • Top 1%≤ 1%
  • Top 5%≤ 5%
  • Top 10%≤ 10%
  • Good≤ 25%
  • Medium≤ 50%
  • Below avg> 50%
Speed
  • Top 1%≥ 345 tok/s
  • Top 5%≥ 237 tok/s
  • Top 10%≥ 196 tok/s
  • Good≥ 146 tok/s
  • Medium≥ 90 tok/s
  • Slow< 90 tok/s
Cost
  • Freeopen weights · self-host
  • Low< $1 / M out
  • Moderate$1–5 / M out
  • High≥ $5 / M out

Why it matters

K2 demonstrates that Chinese open-frontier labs are not just DeepSeek and Qwen — Moonshot, Zhipu, MiniMax, Baichuan, and Stepfun all maintain meaningful open-weight presence in 2025-26. The "Chinese open ecosystem" is a multi-lab phenomenon, not a single-vendor story.

Core Capabilities

Long Documents
Handles entire codebases, books, and multi-doc RAG.
Generative
Produces images, video, audio, or other media.
Agent Workflows
Built for tool use and autonomous tasks.

Context Window

200k tokens
≈ 154+ pages
4k Chat 聊天
32k Long docs 长文档
128k Books 整本书
400k Multi-doc 多文档
1M Codebase 整个代码库
10M
200k

Availability

API
Available
Product / App
Available
Open Source
Released
Enterprise
Contact sales

Pricing Model

Free / self-host
Open weights — pay only for compute.
Self-host

Capability / Performance

Where this model sits relative to the middle 60% of models in the tree. All scores are 0–10 (higher is better).

Lower 20% Upper 80% This model
Reasoning
AA Intelligence Index · scaled to 10
1.7
5.6
6.7
Coding
SciCode · scaled to 10
1.8
4.3
4.9
Agentic tasks
Terminal-Bench Hard · scaled to 10
0.2
3.6
4.4
Context / memory
Context window size · log-scaled
6.0
9.0
6.7
Cost efficiency
Input price ($/M tokens) · cheaper scores higher
6.2
10.0
10.0
Lower 20% 20th percentile — 20% of models score below this This model Where the current model lands Upper 80% 80th percentile — only 20% of models score above this Percentile boundaries are computed across every model in the tree that reports the underlying benchmark for each capability.

What it feels like

  • 1T-parameter MoE / 32B active per token — Moonshot's open-weights debut, modified MIT license
  • 65.8% on SWE-bench Verified single-attempt — outperforms every model tested except Claude Sonnet 4 at release
  • 53.7% on LiveCodeBench v6 — strong open-source coding tier
  • 75.1% on GPQA Diamond — beats GPT-4.1 (66.3%) and Gemini 2.5 Flash (68.2%)
  • Pre-trained on 15.5T tokens with zero training instabilities — the largest stable open MoE training to date
  • Set up for agentic workflows — designed around tool-use, not just chat

Best use cases

  • Self-hosted agent platforms where API models can't go (regulated, private cloud)
  • Cost-sensitive frontier-tier inference via budget providers
  • Coding agents that need long context + open weights for fine-tuning
  • Chinese/multilingual production deployments where Western platforms are blocked

Tools to try

Not ideal for

  • Edge / single-GPU deployments — 1T MoE still demands multi-node serving
  • Multimodal tasks (text-only at this generation; vision lives in Kimi-VL)
  • Latency-sensitive interactive chat at the full model scale

Model Evolution

View full evolution tree →