Kimi K3

kimi/kimi-k3
NewFeatured
VisionToolsJSONReasoning
by MoonshotAI · 2026-07-15

Kimi K3 is Moonshot AI's flagship model and its most capable release to date — a 2.8-trillion-parameter Mixture-of-Experts model built for long-horizon coding and end-to-end knowledge work. It pairs a 1M-token context window with native visual understanding, accepting text and image input with text output, and is designed to stay coherent across very long agentic sessions. Moonshot positions K3 for programming-agent scenarios such as Codex, Claude Code, Cline, and RooCode, as well as deep reasoning and knowledge work. It exposes a top-level reasoning_effort control and native tool calling, and speaks the OpenAI API format for drop-in integration. Note that K3 does not accept sampling parameters (no temperature / top_p / seed) — reasoning depth is controlled through reasoning_effort instead.

ctx1.05M tokens
Inputtext + image
Outputtext
p50 TTFT10.00 s
INPUT$3.00/ 1M tokens
OUTPUT$15.00/ 1M tokens
p50 TTFT10.00 s7d
p95 TTFT10.00 s7d
TRAFFIC849.0Mtokens / 7d

MoonshotAI Kimi K3 is a large language model developed by MoonshotAI, a Chinese AI company. It supports a context window of 1,048,576 tokens and accepts both text and image inputs. The model is…

What is MoonshotAI Kimi K3?

Who is Kimi K3 designed for?

How does Kimi K3 handle multimodal inputs?

What kind of performance can be expected from Kimi K3?

Code samples

Call from any SDK

OpenAI-compatible — keep the SDK you already use

  • OpenAI SDKhttps://api.orcarouter.ai/v1
from openai import OpenAI

client = OpenAI(
    base_url="https://api.orcarouter.ai/v1",
    api_key="$ORCAROUTER_API_KEY",
)

response = client.chat.completions.create(
    model="kimi/kimi-k3",
    messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)

Supported parameters

  • frequency_penalty
  • include_reasoning
  • max_tokens
  • presence_penalty
  • reasoning
  • reasoning_effort
  • response_format
  • stop
  • structured_outputs
  • tool_choice
  • tools

Pricing

Input / 1M tokens$3.00
Output / 1M tokens$15.00
Cache read / 1M$0.300
CurrencyUSD

Cost calculator

Tokens / month10MM
Input share70%%
Estimated / month $66.00 · With prompt caching $56.55

Estimate based on list price

Token & cost estimator

Input tokens: 20Cost per request: $0.007560

Estimate only — actual token counts depend on the provider's tokenizer.

Performance

p50 TTFT
10.00 s
Output speed
54.8 tok/s
p95 TTFT
10.00 s
Error rate
2.4%

Public benchmarks

Source: Design Arena

How it compares

Kimi K3kimi/kimi-k2.6Kimi K2.7 Codekimi/kimi-k2.5
Input $/M$3.00$0.95$0.95$0.60
Output $/M$15.00$4.00$4.00$3.00
Context1.0M262K262K262K
Quality9/108/108/107/10
Compare side-by-sideCompare side-by-sideCompare side-by-sideCompare side-by-side

FAQ

What is the cost to use Kimi K3?
Kimi K3 is priced at $3.00 per 1 million input tokens and $15.00 per 1 million output tokens. This is the provider rate with zero markup. Input tokens include text and image tokens. Output tokens are the generated tokens.
What is the context window size of Kimi K3?
Kimi K3 has a context window of 1,048,576 tokens. This is roughly 1.5 million English words or 800,000 Chinese characters. It can handle very long documents or extensive histories in a single request.
What are the main strengths of Kimi K3?
The main strengths are its very large context window and multimodal support (text and image). It excels at tasks that require processing long sequences, such as full-length book analysis, extensive codebase understanding, and multimodal reasoning across many images.
How does Kimi K3 compare to GPT-4 Turbo?
Kimi K3 has a larger context window (1M vs 128K) and supports images at a lower cost per token. GPT-4 Turbo typically has stronger performance on narrow benchmarks. The best choice depends on your task's context length requirements and quality expectations.
Does OrcaRouter store or train on my data when using Kimi K3?
OrcaRouter does not train on your data. Your prompts are forwarded to the provider (MoonshotAI) for inference. Data handling policies of MoonshotAI apply. OrcaRouter may keep logs for operational purposes but does not use customer data for model improvements.
How do I call Kimi K3 via OrcaRouter's OpenAI-compatible API?
Send requests to https://api.orcarouter.ai/v1/chat/completions with model ID "kimi/kimi-k3". Use standard Chat Completions format with a messages array. For images, use content type "image_url". Include your OrcaRouter API key in the Authorization header.
What is the expected latency for Kimi K3?
Exact latency figures are not provided. Because the model supports very long contexts, processing time increases with input length. For a 1M-token input, expect significantly higher latency than for shorter inputs. Use streaming to get partial responses faster.
Can I use Kimi K3 for real-time applications?
Real-time use is possible but latency may be high for large contexts. It is more suitable for offline batch processing like document analysis. For interactive applications, consider setting a small max_tokens and limiting input length.
What image formats and sizes does Kimi K3 support?
The provided facts do not specify supported image formats or maximum resolution. In practice, most multimodal models accept common formats like JPEG, PNG, and WebP. For best results, use images with moderate resolution and avoid extremely large files to control token cost.
How do I estimate token usage for a given input?
OrcaRouter does not provide a tokenizer, but you can use APIs that return token counts or estimate with libraries like tiktoken (for OpenAI) with caution. For images, assume a fixed token count per image based on resolution. Test with sample inputs to gauge actual usage.

Embed this badge

MoonshotAI: Kimi K3$3.00/M in10000ms p50via OrcaRouter
HTML <a href="https://www.orcarouter.ai/models/kimi/kimi-k3" target="_blank"> <img src="https://www.orcarouter.ai/embed/kimi/kimi-k3.svg" alt="MoonshotAI: Kimi K3 on OrcaRouter" /> </a>
Markdown [![MoonshotAI: Kimi K3](https://www.orcarouter.ai/embed/kimi/kimi-k3.svg)](https://www.orcarouter.ai/models/kimi/kimi-k3)

Model card as data

GET /api/public/models/kimi/kimi-k3Open
Machine-readable:/llms.txt/llms-full.txt