---
title: "Kimi K3 | AI Models | Inference.net"
description: "Run Kimi K3 (2.8T, Moonshot AI) on Inference.net. 1.049M context. tool-calling, structured outputs, vision input, reasoning. Deploy with an OpenAI-compatible API and pay-per-token pricing."
url: "https://inference.net/models/kimi-k3/"
---

Moonshot AI Reasoning Model

# Kimi K3

Moonshot AI's flagship open-weight reasoning model: a 2.8T-parameter Mixture-of-Experts that activates 104B parameters per token, with a 1M-token context window, native vision, and configurable reasoning effort. Frontier-grade agentic coding served on an OpenAI-compatible API at a fraction of frontier pricing.

[Get Started](https://inference.net/signin/) [Documentation](https://docs.inference.net)

Model details

Developed by Moonshot AI

Model family Kimi

Size 2.8T parameters

Architecture Mixture-of-Experts

Context window 1024K tokens

Max output 1024K tokens

License Kimi K3 License

Model API pricing Per 1M tokens

Input $3.00

Cache read $0.30

Output $15.00

[See all pricing](https://inference.net/pricing/)

Kimi K3 is built on a sparse Mixture-of-Experts architecture that activates 104B of its 2.8 trillion parameters per token — routing 16 of 896 experts — keeping latency and cost low while preserving deep reasoning ability. Configurable low, medium, and high reasoning efforts let you trade speed for depth on a per-request basis.

The model supports function calling, structured outputs, and image input, and streams responses over an OpenAI-compatible API.

```
import OpenAI from "openai";

const openai = new OpenAI({
  baseURL: "https://api.inference.net/v1",
  apiKey: "<YOUR_API_KEY>",
});

const completion = await openai.chat.completions.create({
  model: "moonshotai/kimi-k3",
  messages: [
    {
      role: "user",
      content: "What is the meaning of life?"
    }
  ],
  stream: true,
});

for await (const chunk of completion) {
  process.stdout.write(chunk.choices[0]?.delta.content as string);
}
```

Prompt caching

1. No special request field is needed. Keep the prompt prefix stable and byte-identical across calls.
2. Check usage.prompt_tokens_details.cached_tokens in the response to confirm cache reads.

## Related models

[See all models](https://inference.net/models/)

Model

[MiniMax M3.1 Flash 977K input context Vision Tools Reasoning $0.10 / $0.50](https://inference.net/models/minimax-m3.1-flash/)

[Nemotron 3 Super 256K input context Tools $0.102 / $0.48](https://inference.net/models/nemotron-3-super/)

[Schematron V2 Small 125K input context $0.05 / $0.23](https://inference.net/models/schematron-v2-small/)

[Schematron V2 Turbo 125K input context $0.03 / $0.15](https://inference.net/models/schematron-v2-turbo/)

## Start building with Kimi K3 today

Serverless, OpenAI-compatible, and billed per token. No commitments.

[Get Started](https://inference.net/signin/) [Read the docs](https://docs.inference.net)