Kimi K3 API: How to Use Moonshot's 2.8T, 1M-Context Model (OpenAI-Compatible)

Moonshot AI announced Kimi K3 on July 16, 2026. This guide covers what it is, honest pricing, and copy-paste code to call it today through an OpenAI-compatible endpoint.

What is Kimi K3?

Kimi K3 is the newest flagship in Moonshot AI's Kimi K-series. Per Moonshot's launch announcement and reporting around it, K3 is:

• A 2.8-trillion-parameter Mixture-of-Experts (MoE) model — described as the first open model in the "3T class".
• Natively multimodal (vision + text).
• Built around a 1-million-token context window.
• Planned to ship with open weights (announced for late July 2026).

It sits above the earlier K-series releases (K2.5, K2.6, and the K2.7 coding variants), which remain available for cost- or latency-sensitive workloads.

Kimi K3 API Pricing (Honest Numbers)

Here is the real per-token pricing on NovAI, next to OpenRouter's list price for the same model. We don't claim to be the cheapest everywhere — for K3 the difference is small:

Rule of thumb: use K3 when you genuinely need frontier multimodal reasoning or the full 1M context; drop to K2.5 when you want long context at a fraction of the cost. Both share the same API — you only change the model string.

How to Call Kimi K3 (OpenAI-Compatible)

NovAI exposes Kimi K3 on a standard OpenAI-compatible endpoint, so any OpenAI SDK works. You change two things: the base_url and the model name.

curl

curl https://aiapi-pro.com/v1/chat/completions \
  -H "Authorization: Bearer $NOVAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kimi-k3",
    "messages": [
      {"role": "user", "content": "Summarize the key risks in this 200-page contract."}
    ]
  }'

Python (openai SDK)

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_NOVAI_KEY",
    base_url="https://aiapi-pro.com/v1",
)

resp = client.chat.completions.create(
    model="kimi-k3",
    messages=[
        {"role": "user", "content": "Explain MoE routing in 3 bullet points."}
    ],
)
print(resp.choices[0].message.content)

Node.js

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.NOVAI_API_KEY,
  baseURL: "https://aiapi-pro.com/v1",
});

const resp = await client.chat.completions.create({
  model: "kimi-k3",
  messages: [{ role: "user", content: "Give me a haiku about long context." }],
});
console.log(resp.choices[0].message.content);

Vision (image + text)

Because K3 is multimodal, you can pass images in the same request format the OpenAI SDK uses:

resp = client.chat.completions.create(
    model="kimi-k3",
    messages=[{
        "role": "user",
        "content": [
            {"type": "text", "text": "What's in this chart?"},
            {"type": "image_url", "image_url": {"url": "https://example.com/chart.png"}}
        ]
    }],
)

When Kimi K3 Makes Sense

• Long-document / whole-repo tasks — 1M tokens fits large contracts, books, or codebases in one call.
• Multimodal reasoning — mixing images and text without a separate vision model.
• Frontier-quality output where the K2.5/K2.6 tiers aren't enough.

For high-volume, cost-sensitive workloads (classification, extraction, routing), a cheaper model on the same endpoint — Kimi K2.5, DeepSeek-V4-Flash, or Qwen — will usually be the better economic choice.

Disclosure: NovAI (aiapi-pro.com) is an independent, OpenAI-compatible API gateway. We resell access to Kimi K3 and other models; we are not Moonshot AI. Model specifications above are from Moonshot's public announcement; pricing is our live list price as of July 2026.

Frequently Asked Questions

What is Kimi K3?

Kimi K3 is Moonshot AI's flagship model announced on July 16, 2026 — a 2.8-trillion-parameter MoE model with native vision and a 1-million-token context window, with open weights planned. It is the newest entry in the Kimi K-series after K2.5, K2.6, and K2.7.

How do I call the Kimi K3 API?

Point your OpenAI SDK at https://aiapi-pro.com/v1 and set the model to kimi-k3. Your existing chat-completions code, streaming, and tool calling work unchanged — you only swap the base URL and model name.

How much does the Kimi K3 API cost?

On NovAI, Kimi K3 is $2.90 per 1M input tokens and $14.50 per 1M output tokens with the full 1M context. OpenRouter lists it at $3.00 / $15.00. If you need lower cost, Kimi K2.5 ($0.64 / $3.35) is a cheaper long-context option on the same API.

What is the Kimi K3 context window?

1 million tokens — suited to long-document analysis, large codebases, and multi-file reasoning in a single request.

Is Kimi K3 multimodal?

Yes. Moonshot describes K3 as natively multimodal with vision, so it can reason over images alongside text within the same 1M-token context.