Kimi K3

Kimi's flagship reasoning model for long-horizon coding, visual understanding, and end-to-end knowledge work.

1M ContextNative VisionAgentic Tool Use

Context Window

1,048,576 tokens

kimi-k3

Max Output

1,048,576 tokens

1M tokens

Input

Output

$13.5$15

10% off

1M tokens

Model Codekimi-k3

Playground

Model Code

kimi-k3

Reference inputs

Reference Images0/5

Click or drag

Enter your prompt here...

Output

No output yet

Pricing

Plan	Context Window	Max Output	Input	Output	Cache Write	Cache Read
kimi-k3	1,048,576 tokens	1,048,576 tokens	-	$13.5$1510% off 1M tokens	-	$0.27$0.3010% off 1M tokens

Cache Miss Input · $2.7 · 1M tokens

Cache read or cached input pricing applies to reused prompt tokens.

README

Kimi K3 is Kimi's most capable flagship model, combining always-on reasoning, native visual understanding, a 1M-token context window, and advanced tool use for long-horizon coding and knowledge work.

Kimi K3 is a 2.8-trillion-parameter flagship model built for long-horizon coding, end-to-end knowledge work, reasoning, and workflows that combine software engineering with visual feedback.
The model has native visual understanding and a 1,048,576-token context window. Its maximum completion length can be configured up to 1,048,576 tokens for unusually long reasoning or generation tasks.
Kimi K3 always reasons and supports low, high, and max reasoning effort, with max as the default. It also supports automatic context caching, tool calls, tool-choice constraints, dynamic tool loading, JSON Mode, strict structured output, Partial Mode, and streaming.

API

Use the API docs for request format, parameters, and integration details.

Kimi's flagship reasoning model for long-horizon coding, visual understanding, and end-to-end knowledge work.

Model Code

kimi-k3

View API Docs

API

Endpoint

POST /v1/messages

Authentication

Authorization: Bearer YOUR_REACH_API_KEY
anthropic-version: 2023-06-01
Content-Type: application/json

Example

curl -X POST "https://direct.reachapi.ai/v1/messages" \
  -H "Authorization: Bearer YOUR_REACH_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "YOUR_MODEL_ID",
    "messages": [
      {
        "role": "user",
        "content": "Introduce ReachAPI in one sentence."
      }
    ],
    "max_tokens": 512,
    "temperature": 0.2
  }'

Related Models

Other models from the same vendor

GLM-5.2

Z.AI's GLM-5.2 model for enterprise text generation, reasoning, and OpenAI-compatible API workloads.

DeepSeek-V4-Flash

DeepSeek's fast V4 model for cost-sensitive, OpenAI-compatible text and reasoning workloads.

DeepSeek-V4-Pro

DeepSeek's stronger V4 model for OpenAI-compatible reasoning, coding, and enterprise text workloads.