NOUMENOpen app
NOUMEN

Moonshot · Open weights

Kimi K2 0905

Added Sep 25, 2025Open weights

Compare models

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

Loading models…

Use in the app

Kimi K2 0905. Kimi-k2 is a Mixture-of-Experts (MoE) foundation model with exceptional coding and agent capabilities, featuring 1 trillion total parameters and 32 billion activated parameters. In benchmark evaluations covering general knowledge reasoning, programming, mathematics, and agent-related tasks, the K2 model outperforms other leading open-source models. Quantized at FP8.

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

Pricing

USD per 1M tokens
Input
$0.40
Output
$1.80

Reference prices as published by the model provider. API example

Specifications

Context window
262.1K
Max output
100.4K
Average output (7 days)
677 tokens
Longer than 60% of 465 models

Side by side

Kimi K2 0905 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 0905MoonshotThis modelQwen3.6 35B A3B TEEQwenRemove
ProviderMoonshotQwen
Discovery tierOpen weightsOpen weights
AddedSep 25, 2025Jun 9, 2026
Limits
Context window262.14K262.14K
Max output100.35K262.14KBest
Average output677 tokens3.9K tokens
Pricing, USD per 1M tokens
Input$0.40$0.20Best
Output$1.80$1.27Best
Performance
Throughput50.8 tok/s—
Time to first token3.3 s—
Request success (24h)100.0%Best99.5%
Benchmark indexes
Intelligence Index15.3Best15.2
Coding Index—28.1
Math Index57.3—
Capabilities
ToolsYesNo
Structured outputYesNo
Reasoning effort——

Public evaluations

Benchmarks

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

  • Kimi K2 0905
  • Qwen3.6 35B A3B TEE

Intelligence Index

  • Kimi K2 090515.3
  • Qwen3.6 35B A3B TEE15.2

Kimi K2 0905: Better than 39% of 326 scored models

3.8Catalog models by score55.1

Indexes

  • Coding Index

    Composite of the coding evaluations

    • Kimi K2 0905: —
    • Qwen3.6 35B A3B TEE: 28.1

  • Math Index

    Composite of the math evaluations

    • Kimi K2 0905: 57.3
    • Qwen3.6 35B A3B TEE: —

Agentic work

  • GDPval-AAv2.1

    Economically valuable occupational tasks, rated head to head

    • Kimi K2 0905: —
    • Qwen3.6 35B A3B TEE: 847 Elo

Document reasoning

  • AA-LCRv1.1

    Reasoning across long documents

    • Kimi K2 0905: 53.0%
    • Qwen3.6 35B A3B TEE: 64.3%Top

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • Kimi K2 0905: 6.4%
    • Qwen3.6 35B A3B TEE: 13.9%Top

  • GPQA Diamond

    Graduate-level science questions

    • Kimi K2 0905: 76.7%
    • Qwen3.6 35B A3B TEE: 81.7%Top

  • MMLU-Pro

    Multi-subject knowledge and reasoning

    • Kimi K2 0905: 81.9%
    • Qwen3.6 35B A3B TEE: —

  • IFBench

    Following precise output instructions

    • Kimi K2 0905: 41.7%Top
    • Qwen3.6 35B A3B TEE: 36.2%

Coding

  • LiveCodeBench

    Recent competitive programming problems

    • Kimi K2 0905: 61.0%
    • Qwen3.6 35B A3B TEE: —

Math

  • AIME 2025

    Competition mathematics, 2025 problems

    • Kimi K2 0905: 57.3%
    • Qwen3.6 35B A3B TEE: —

Legacy evaluations

  • Terminal-Bench Hard

    Earlier terminal task suite

    • Kimi K2 0905: 23.5%
    • Qwen3.6 35B A3B TEE: 25.8%Top

  • τ²-Bench Telecom

    Tool use in a customer-support simulation

    • Kimi K2 0905: 73.4%
    • Qwen3.6 35B A3B TEE: 85.1%Top

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • Kimi K2 0905: 53.0%
    • Qwen3.6 35B A3B TEE: 64.3%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
50.8 tok/s
Faster than 38% of 250 models
Time to first token
3.3 s
Quicker than 52% of 250 models
Prompt cache hits
67.5%
Request success (24h)
100.0%
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 100.0%
  • 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-Instruct-0905",
    "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-Instruct-0905",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

Moonshot

Other Moonshot models, starting with the same family.