Moonshot · Open weights
Kimi K2 Instruct
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
- 256K
- Max output
- 8.2K
- Average output (7 days)
- 564 tokens
- Longer than 53% of 465 models
Side by side
Kimi K2 Instruct compared
Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.
| Kimi K2 InstructMoonshotThis model | Kimi K2.6 ThinkingMoonshotRemove | |
|---|---|---|
| Provider | Moonshot | Moonshot |
| Discovery tier | Open weights | Open weights |
| Added | Jan 1, 2024 | Apr 16, 2026 |
| Limits | ||
| Context window | 256K | 262.14KBest |
| Max output | 8.19K | 65.54KBest |
| Average output | 564 tokens | 4K tokens |
| Pricing, USD per 1M tokens | ||
| Input | $0.40Best | $0.45 |
| Output | $1.80Best | $1.89 |
| Cache read | — | $0.0757 |
| Performance | ||
| Throughput | — | 40.8 tok/s |
| Time to first token | — | 2.9 s |
| Request success (24h) | — | 99.8% |
| Benchmark indexes | ||
| Intelligence Index | 12.7 | 27.0Best |
| Coding Index | — | 61.8 |
| Math Index | 57.0 | — |
| Capabilities | ||
| Reasoning | No | Yes |
| Tools | Yes | Yes |
| Vision | No | Yes |
| Structured output | Yes | No |
| Parallel tools | No | Yes |
| Reasoning effort | — | — |
Public evaluations
Benchmarks
Independent evaluation results, with how each score ranks among the models in the Noumen catalog.
- Kimi K2 Instruct
- Kimi K2.6 Thinking
Intelligence Index
- Kimi K2 Instruct12.7
- Kimi K2.6 Thinking27.0
Kimi K2 Instruct: Better than 28% of 326 scored models
Indexes
Coding Index
Composite of the coding evaluations
- Kimi K2 Instruct: —
- Kimi K2.6 Thinking: 61.8
Math Index
Composite of the math evaluations
- Kimi K2 Instruct: 57.0
- Kimi K2.6 Thinking: —
Agentic work
AutomationBench-AA
Workflow automation, penalized for guardrail violations
- Kimi K2 Instruct: —
- Kimi K2.6 Thinking: 13.0%
AutomationBench-AA tasks completed
Workflows finished end to end without a guardrail violation
- Kimi K2 Instruct: —
- Kimi K2.6 Thinking: 1.8%
AA-Briefcasev1.1
Agentic knowledge work, rated head to head
- Kimi K2 Instruct: —
- Kimi K2.6 Thinking: 819 Elo
GDPval-AAv2.1
Economically valuable occupational tasks, rated head to head
- Kimi K2 Instruct: —
- Kimi K2.6 Thinking: 1,038 Elo
Document reasoning
GDP.pdf
Professional reasoning over PDF documents, all-pass rate
- Kimi K2 Instruct: —
- Kimi K2.6 Thinking: 13.0%
AA-LCRv1.1
Reasoning across long documents
- Kimi K2 Instruct: 53.0%
- Kimi K2.6 Thinking: 81.0%Top
MLCR-AA
Long-context reasoning over medical records
- Kimi K2 Instruct: —
- Kimi K2.6 Thinking: 11.1%
Reasoning and knowledge
Humanity’s Last Exam
Expert-level questions across academic fields
- Kimi K2 Instruct: 7.4%
- Kimi K2.6 Thinking: 37.5%Top
GPQA Diamond
Graduate-level science questions
- Kimi K2 Instruct: 76.6%
- Kimi K2.6 Thinking: 91.1%Top
MMLU-Pro
Multi-subject knowledge and reasoning
- Kimi K2 Instruct: 82.4%
- Kimi K2.6 Thinking: —
IFBench
Following precise output instructions
- Kimi K2 Instruct: 41.5%
- Kimi K2.6 Thinking: 76.0%Top
Coding
Terminal-Benchv4.0
Practical software engineering in a terminal
- Kimi K2 Instruct: —
- Kimi K2.6 Thinking: 0.5%
SciCode
Scientific computing in Python
- Kimi K2 Instruct: —
- Kimi K2.6 Thinking: 51.5%
LiveCodeBench
Recent competitive programming problems
- Kimi K2 Instruct: 55.6%
- Kimi K2.6 Thinking: —
Math
AIME 2025
Competition mathematics, 2025 problems
- Kimi K2 Instruct: 57.0%
- Kimi K2.6 Thinking: —
AIME
Competition mathematics
- Kimi K2 Instruct: 69.3%
- Kimi K2.6 Thinking: —
MATH-500
Competition math problems
- Kimi K2 Instruct: 97.1%
- Kimi K2.6 Thinking: —
Legacy evaluations
Terminal-Bench Hard
Earlier terminal task suite
- Kimi K2 Instruct: 15.9%
- Kimi K2.6 Thinking: 43.9%Top
τ²-Bench Telecom
Tool use in a customer-support simulation
- Kimi K2 Instruct: 61.1%
- Kimi K2.6 Thinking: 95.9%Top
AA-LCR (unversioned)
Long-context reasoning, earlier grading
- Kimi K2 Instruct: 53.0%
- Kimi K2.6 Thinking: 81.0%Top
Scores from Artificial Analysis, updated Oct 26, 2026. Artificial Analysis
- Kimi K2 Instruct
- Kimi K2.6 Thinking
| Leaderboard | Kimi K2 Instruct | Kimi K2.6 Thinking |
|---|---|---|
| EQ-Bench 3 | #11 · 1548 | Not in top 20 |
Ranks among Noumen catalog models on EQ-Bench leaderboards. Only top-20 placements are listed.EQ-Bench
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.
- Prompt cache hits
- 67.0%
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.
Developers
Use it through the API
Noumen’s API is OpenAI-compatible: any Chat Completions client works with a base URL and a key.
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",
"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",
messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)Moonshot
Related models
Other Moonshot models, starting with the same family.