NOUMENOpen app
NOUMEN

Moonshot · Open weights

Kimi K2 Thinking

Added Nov 6, 2025Open weights

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Use in the app

Moonshot AI's distinct Kimi K2 Thinking checkpoint is a mandatory-reasoning model for long-horizon agentic workflows and multi-step tool use.

  • Reasoning
  • Tools
  • Structured output
Input
Text
Output
Text
Discovery tier
Open weights

Pricing

USD per 1M tokens
Input
$0.60
Output
$2.50
Cache read
$0.15

Reference prices as published by the model provider. API example

Specifications

Context window
262.1K
Max output
98.3K
Average output (7 days)
1.9K tokens
Longer than 91% of 465 models

Side by side

Kimi K2 Thinking compared

Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Kimi K2 ThinkingMoonshotThis modelMiMo V2.5 ThinkingxiaomiRemove
ProviderMoonshotxiaomi
Discovery tierOpen weightsOpen weights
AddedNov 6, 2025Jun 3, 2026
Limits
Context window262.14K1.05MBest
Max output98.3K131.07KBest
Average output1.9K tokens1.1K tokens
Pricing, USD per 1M tokens
Input$0.60$0.14Best
Output$2.50$0.28Best
Cache read$0.15$0.0028Best
Cache write—$0.00
Performance
Throughput48.0 tok/sBest29.4 tok/s
Time to first token2.6 sBest3.8 s
Request success (24h)99.6%100.0%Best
Benchmark indexes
Intelligence Index22.0Best21.8
Coding Index—56.8
Math Index94.7—
Capabilities
ReasoningYesYes
ToolsYesYes
VisionNoYes
Structured outputYesYes
Audio inputNoYes
Video inputNoYes
Reasoning effort—none, high

Public evaluations

Benchmarks

Independent evaluation results, with how each score ranks among the models in the Noumen catalog.

  • Kimi K2 Thinking
  • MiMo V2.5 Thinking

Intelligence Index

  • Kimi K2 Thinking22.0
  • MiMo V2.5 Thinking21.8

Kimi K2 Thinking: Better than 53% of 326 scored models

3.8Catalog models by score55.1

Indexes

  • Coding Index

    Composite of the coding evaluations

    • Kimi K2 Thinking: —
    • MiMo V2.5 Thinking: 56.8

  • Math Index

    Composite of the math evaluations

    • Kimi K2 Thinking: 94.7
    • MiMo V2.5 Thinking: —

Agentic work

  • AutomationBench-AA

    Workflow automation, penalized for guardrail violations

    • Kimi K2 Thinking: —
    • MiMo V2.5 Thinking: 18.4%

  • AutomationBench-AA tasks completed

    Workflows finished end to end without a guardrail violation

    • Kimi K2 Thinking: —
    • MiMo V2.5 Thinking: 3.3%

  • AA-Briefcasev1.1

    Agentic knowledge work, rated head to head

    • Kimi K2 Thinking: —
    • MiMo V2.5 Thinking: 754 Elo

  • GDPval-AAv2.1

    Economically valuable occupational tasks, rated head to head

    • Kimi K2 Thinking: —
    • MiMo V2.5 Thinking: 996 Elo

Document reasoning

  • GDP.pdf

    Professional reasoning over PDF documents, all-pass rate

    • Kimi K2 Thinking: —
    • MiMo V2.5 Thinking: 4.0%

  • AA-LCRv1.1

    Reasoning across long documents

    • Kimi K2 Thinking: 72.0%
    • MiMo V2.5 Thinking: 73.0%Top

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • Kimi K2 Thinking: 23.8%
    • MiMo V2.5 Thinking: 27.2%Top

  • GPQA Diamond

    Graduate-level science questions

    • Kimi K2 Thinking: 83.8%
    • MiMo V2.5 Thinking: 84.9%Top

  • MMLU-Pro

    Multi-subject knowledge and reasoning

    • Kimi K2 Thinking: 84.8%
    • MiMo V2.5 Thinking: —

  • IFBench

    Following precise output instructions

    • Kimi K2 Thinking: 68.1%Top
    • MiMo V2.5 Thinking: 67.1%

Coding

  • Terminal-Benchv4.0

    Practical software engineering in a terminal

    • Kimi K2 Thinking: —
    • MiMo V2.5 Thinking: 0.0%

  • SciCode

    Scientific computing in Python

    • Kimi K2 Thinking: —
    • MiMo V2.5 Thinking: 43.9%

  • LiveCodeBench

    Recent competitive programming problems

    • Kimi K2 Thinking: 85.3%
    • MiMo V2.5 Thinking: —

Math

  • AIME 2025

    Competition mathematics, 2025 problems

    • Kimi K2 Thinking: 94.7%
    • MiMo V2.5 Thinking: —

Legacy evaluations

  • Terminal-Bench Hard

    Earlier terminal task suite

    • Kimi K2 Thinking: 31.1%
    • MiMo V2.5 Thinking: 41.7%Top

  • τ²-Bench Telecom

    Tool use in a customer-support simulation

    • Kimi K2 Thinking: 93.0%Top
    • MiMo V2.5 Thinking: 90.6%

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • Kimi K2 Thinking: 72.0%
    • MiMo V2.5 Thinking: 73.0%Top

Scores from Artificial Analysis, updated Oct 26, 2026. Artificial Analysis

Live routing data

Performance and reliability

Recent averages across requests routed for this model. Your own speed depends on prompt length, reasoning effort and load.

Throughput
48.0 tok/s
Faster than 36% of 250 models
Time to first token
2.6 s
Quicker than 64% of 250 models
Prompt cache hits
28.9%
About $0.47 per 1M input tokens with caching
Request success (24h)
99.6%
Reliable

Cache hits are the share of input tokens recently served from a prompt cache across all requests. They are not a forecast of your own hit rate.

Hourly request success · last 72 hoursLowest hour 95.4%
  • Healthy (98% or more)
  • Degraded (90–98%)
  • Down (under 90%)
  • No requests

Developers

Use it through the API

Noumen’s API is OpenAI-compatible: any Chat Completions client works with a base URL and a key.

Read the quickstart
curl https://www.noumen.xyz/api/v1/chat/completions \
  -H "Authorization: Bearer $NOUMEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "moonshotai/kimi-k2-thinking",
    "messages": [{ "role": "user", "content": "Explain this stack trace." }]
  }'
import os
from openai import OpenAI

client = OpenAI(base_url="https://www.noumen.xyz/api/v1", api_key=os.environ["NOUMEN_API_KEY"])
reply = client.chat.completions.create(
    model="moonshotai/kimi-k2-thinking",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

Moonshot

Other Moonshot models, starting with the same family.