What makes Kimi K3 notable
Kimi K3 is a 2.8-trillion-parameter model that pushes the boundary of what open models can do. It combines extreme Mixture-of-Experts sparsity with a hybrid attention architecture (Kimi Delta Attention + Attention Residuals). Roughly three-quarters of its layers use an efficient linear-attention-style mechanism, while a smaller number of full-attention layers preserve strong global recall.
It ships with:
• A native 1-million-token context window
• Strong multimodal capabilities (text, images, and video) powered by a dedicated vision encoder
• Excellent performance on long-horizon coding, knowledge work, and complex reasoning tasks
One of the most practical advantages is its native MXFP4 quantization. The model was trained with 4-bit weights from the supervised fine-tuning stage onward, which dramatically reduces memory footprint while keeping accuracy high. This makes high-quality inference far more accessible than a traditional BF16 checkpoint of the same scale.
Why Choose Canopy Wave's Shared Endpoint
Running a model of Kimi K3's scale yourself requires significant GPU resources and operational effort. Canopy Wave's Shared Endpoint removes that complexity while giving you production-ready access:
• Fully OpenAI-compatible — Switch from other providers by changing only the base_url and API key. Existing code, tools, and SDKs work with almost no modification.
• Zero infrastructure management — No need to provision GPUs, handle scaling, or worry about cold starts. Just call the API.
• Enterprise-grade reliability and security — Built for production workloads with high availability, strong isolation, and privacy-focused design (including zero data retention policies where applicable).
• Simple pay-as-you-go pricing — You only pay for the tokens you use. No idle costs.
• Fast and stable performance — Optimized inference stack designed for low latency and consistent throughput, even under concurrent load.
• Easy path to scale — Start on the Shared Endpoint and move to a Dedicated Endpoint when you need higher throughput or guaranteed capacity.
This combination makes Canopy Wave one of the simplest and most reliable ways to use Kimi K3 in real applications.
How to get started with Kimi K3 on Canopy Wave
Canopy Wave's Shared Endpoint is fully OpenAI-compatible. You can use the exact same tools and libraries you already know.
Prerequisites
• Set your key as an environment variable:
export CANOPYWAVE_API_KEY="your_api_key_here"
• curl (for the first example)
• Python 3.10 or higher, and the openai package (for the Python examples)
1. Quick test with cURL
curl -X POST https://inference.canopywave.io/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $CANOPYWAVE_API_KEY" \
-d '{
"model": "moonshotai/kimi-k3",
"messages": [
{"role": "user", "content": "tell me a story"}
],
"max_tokens": 1000,
"temperature": 0.7
}'2. Using the OpenAI Python client (recommended)
from openai import OpenAI
import os
BASE_URL = "https://inference.canopywave.io/v1"
API_KEY = os.environ.get("CANOPYWAVE_API_KEY")
client = OpenAI(api_key=API_KEY, base_url=BASE_URL)
response = client.chat.completions.create(
model="moonshotai/kimi-k3",
messages=[
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "please tell me a story."}
],
)
print(response.choices[0].message.content)Tips for getting the most out of Kimi K3
• Use a clear system prompt when you need specific behavior (coding style, role-playing, etc.).
• Take advantage of the full 1M context window for large codebases, long documents, or multi-turn agent workflows.
• The model ID must be exactly moonshotai/kimi-k3.
Conclusion
Kimi K3 is one of the most capable open models available today for advanced coding, long-context knowledge work, and reasoning. With Canopy Wave's Shared Endpoint, you can start using it immediately — no cluster to manage, no cold starts, and full OpenAI compatibility.
Frequently Asked Questions
Q1: What is the exact model ID I should use?
A: Always use moonshotai/kimi-k3.
Q2: Is the API fully compatible with the OpenAI Python/JS SDK?
A: Yes. Just set the base_url to https://inference.canopywave.io/v1 and use your Canopy Wave API key.
Q3: Does Kimi K3 on Canopy Wave support the full 1-million-token context?
A: Yes, the Shared Endpoint supports the model's native 1M context window.
Q4: How is pricing calculated?
A: Pricing is based on input tokens, output tokens, and cache hits. You can find the latest rates on our Pricing page.
Q5: Which tools are compatible with Canopy Wave's API?
A: We provide OpenAI-compatible APIs that work with many popular tools, including Cline, Roo Code, Kilo Code, Dify, OpenClaw, Open Code, Silly Tavern, Cherry Studio, Chatbox, Open WebUI, and more. Just set the base URL to https://inference.canopywave.io/v1 and use your Canopy Wave API key.
