kimi/kimi-k3
1,048,576 context · $1.80 / M input tokens · $9.00 / M output tokens
Kimi K3 is a 2.8-trillion-parameter multimodal reasoning model described for long-horizon coding, knowledge work, visual understanding, and agentic workflows. Use the verified kimi-k3 route through OurToken's unified API.
2026-07-17 06:46:29 UTC
Pricing
Pay-per-use
No upfront costs, pay only for what you use
API Usage
API Access Guide
Code examples
Use the OurToken API endpoint for this model. The examples below use direct HTTP requests and the recommended endpoint for the model family.
curl https://api.ourtoken.ai/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"model": "kimi-k3",
"messages": [
{
"role": "user",
"content": "Hello!"
}
],
"max_tokens": 256
}'Chat Completions API Reference
Create a chat response with the OpenAI Chat Completions-compatible endpoint. Use https://api.ourtoken.ai/v1 as the SDK Base URL and POST /chat/completions as the endpoint.
Authorization
| Content-Type | application/json |
| Authorization | Bearer YOUR_API_KEY |
Request Body
| Field | Type | Required | Description |
|---|---|---|---|
| model | string | Required | Model ID to call. |
| messages | array<object> | Required | Conversation messages sent to the model. |
| max_tokens | integer | Optional | Maximum number of output tokens. |
| temperature | number | Optional | Sampling temperature. |
| top_p | number | Optional | Nucleus sampling parameter. |
| stream | boolean | Optional | Whether to return a streaming response. |
| stream_options | object | Optional | Additional options for streaming responses. |
| tools | array<object> | Optional | Tools available to the model. |
| tool_choice | string | object | Optional | Controls how the model selects tools. |
| response_format | object | Optional | Controls structured output, such as JSON object responses. |
Response Body
| Field | Type | Required | Description |
|---|---|---|---|
| id | string | Required | Unique chat completion identifier. |
| object | "chat.completion" | Required | Object type returned by the Chat Completions API. |
| created | integer | Required | Unix timestamp when the response was created. |
| model | string | Required | Model that produced the response. |
| choices | array<object> | Required | Candidate responses returned by the model. |
| choices[].message.role | string | Required | Role of the returned chat message. |
| choices[].message.content | string | Optional | Text content in the returned chat message. |
| choices[].finish_reason | string | Optional | Reason generation stopped. |
| usage | object | Optional | Token usage information for the chat completion. |
| usage.prompt_tokens | integer | Optional | Input token count. |
| usage.completion_tokens | integer | Optional | Output token count. |
| usage.total_tokens | integer | Optional | Total token count. |
| usage.prompt_tokens_details | object | Optional | Breakdown of input token usage. |
| usage.prompt_tokens_details.cached_tokens | integer | Optional | Tokens served from cache. |
Model Introduction
Kimi kimi-k3
Kimi K3 is a 2.8-trillion-parameter multimodal reasoning model described for long-horizon coding, knowledge work, visual understanding, and agentic workflows. Use the verified kimi-k3 route through OurToken's unified API.
Kimi K3 is Kimi's flagship model with a 1,048,576-token context window and native visual understanding described in the official Kimi materials. Use Kimi K3 API through OurToken to test coding, knowledge-work, repository, tool, image, and runtime-feedback workflows with one API key and one model ID.
Why It Looks Great
- Kimi K3 API access through a unified OurToken endpoint.
- 1,048,576-token context window for long documents, repositories, logs, and task history.
- Native visual understanding described for coding, frontend, game development, CAD, and related workflows.
- Current platform pricing at 60% of the supplied official reference prices.
- Official materials describe long-horizon coding and knowledge-work use cases.
Key Features
- Model ID: kimi-k3
- Context window: 1,048,576 tokens
- Parameters: 2.8T
- OurToken input price: $1.80 / 1M tokens
- OurToken output price: $9.00 / 1M tokens
- OurToken cache-hit price: $0.18 / 1M tokens
- Provider: Kimi
Specifications
Kimi K3 API Features for Coding and Long-Context Work
Explore the Kimi K3 API through a unified developer workflow with transparent Kimi K3 pricing, a 1M-token context window, visual understanding, and long-horizon coding support.
Unified API Access
Use the Kimi K3 API through OurToken with a consistent model ID and unified API workflow. Developers can manage access through the same platform used for other supported models, making it easier to test Kimi K3 in existing applications and agent workflows.
Transparent Model Pricing
The Kimi K3 price on OurToken is shown alongside the official reference price, with input, cache-hit, and output rates separated clearly. The Kimi K3 pricing structure uses the confirmed 60% platform rate, while Cache Write remains undisplayed because no separate official price was provided.
Million-Token Context
Kimi K3 provides a 1,048,576-token context window for tasks that require large inputs, long documents, or extended project history. This context capacity is useful when a Kimi K3 model needs to work across substantial codebases, logs, specifications, and iterative task state.
Long-Horizon Coding
Kimi K3 is designed for long-horizon coding workflows that involve understanding large repositories, coordinating terminal tools, debugging, and iterating from runtime feedback. These workflows can help developers explore complex projects and sustain multi-step engineering tasks with less manual context switching.
Visual Reasoning Workflows
The official Kimi K3 description includes native visual understanding and visual feedback workflows. This makes the model relevant to frontend engineering, game development, CAD, and debugging tasks where screenshots or other visual signals need to inform the next engineering step.
Open-Weight Frontier Model
Moonshot Kimi K3 is presented as a 2.8-trillion-parameter open-source or open-weight multimodal reasoning model. Its KDA, Attention Residuals, and sparse expert architecture are part of the official model positioning, while detailed benchmark claims should wait for the technical report.
How to Use the Kimi K3 API on OurToken
Follow six practical steps to create an API key, select the Kimi K3 model ID, send requests, review Kimi K3 API pricing, and monitor usage.
Create an API Key
Sign in to OurToken and create an API key from the API key management page. Keep the key private, assign only the required model access, and use a separate key for development, testing, or production workflows.
01Select the Model ID
Set the requested model to `kimi-k3`, which is the verified model ID for this integration. If a client displays Moonshot Kimi K3 or another provider label, confirm that the outgoing request still uses the exact `kimi-k3` identifier.
02Send an API Request
Configure your client with the OurToken API endpoint and the new API key, then send a standard chat request using `kimi-k3`. Start with a small test prompt before moving the model into longer coding, knowledge-work, or agentic workflows.
03Review Kimi K3 Pricing
Check the Kimi K3 API pricing before scaling usage. OurToken lists input, cache-hit, and output prices separately, with the current platform rates shown against the official reference so you can estimate request costs more clearly.
04Test Engineering Workflows
Evaluate Kimi K3 with a representative coding or knowledge-work task. For long-horizon work, provide repository context gradually, inspect tool or runtime feedback, and verify that the model handles the actual task structure before increasing request size.
05Monitor Usage and Cost
Review request history, token usage, model-level costs, and remaining API budget in the OurToken dashboard. Compare real usage with your estimate, then adjust prompts, request length, or key limits before production traffic grows.
06Kimi K3 FAQ
Find direct answers about the Kimi K3 API, Kimi K3 pricing, model ID, release timing, open-source status, coding workflows, and official benchmark information.