kimi-k3
kimi-k3
1,002<br>Downloads
Updated<br>an hour ago
Kimi K3 is an open-weight, native multimodal agentic model and our most capable model to date.
Kimi K3 is an open-weight, native multimodal agentic model and our most capable model to date.
Cancel
vision
tools
thinking
cloud
Usage
extra high
Context
1M
tokens
Size
2.81T
parameters
CLI
cURL<br>Python<br>JavaScript
Documentation<br>Documentation
ollama run kimi-k3:cloud
curl http://localhost:11434/api/chat \<br>-d '{<br>"model": "kimi-k3:cloud",<br>"messages": [{"role": "user", "content": "Hello!"}]<br>}'
from ollama import chat
response = chat(<br>model='kimi-k3:cloud',<br>messages=[{'role': 'user', 'content': 'Hello!'}],<br>print(response.message.content)
import ollama from 'ollama'
const response = await ollama.chat({<br>model: 'kimi-k3:cloud',<br>messages: [{role: 'user', content: 'Hello!'}],<br>})<br>console.log(response.message.content)
Applications
Claude Code<br>ollama launch claude --model kimi-k3:cloud
OpenCode<br>ollama launch opencode --model kimi-k3:cloud
Hermes Agent<br>ollama launch hermes --model kimi-k3:cloud
OpenClaw<br>ollama launch openclaw --model kimi-k3:cloud
Models
View all →
Name
1 model
Size / Usage
Context
Input
kimi-k3:cloud
Extra High Usage · 1M context window · Text, Image · an hour ago
kimi-k3:cloud
1M
Text, Image
Readme
Currently Kimi K3 requires a Pro or Max subscription, and consumes extra usage credits . We’re quickly working on adding capacity to expand access.
Kimi K3 is an open-weight, native multimodal agentic model and our most capable model to date. It is a 2.8T-parameter model built on Kimi Delta Attention (KDA) and Attention Residuals (AttnRes), with native vision capabilities and a 1-million-token context window. It is the world’s first open 3T-class model, designed for frontier intelligence across long-horizon coding, knowledge work, and reasoning.
Key Features
New Architecture : Kimi K3 is built on Kimi Delta Attention (KDA) and Attention Residuals (AttnRes), and scales up MoE sparsity with a Stable LatentMoE framework that activates 16 out of 896 experts — yielding an approximate 2.5× improvement in overall scaling efficiency over Kimi K2.
Long-Horizon Coding : Operating with minimal human oversight, Kimi K3 sustains long engineering sessions, navigates massive repositories, and orchestrates terminal tools — from GPU kernel optimization and compiler development to vision-in-the-loop game dev, CAD, and even chip design.
Agentic Knowledge Work : Kimi K3 advances end-to-end knowledge work, producing deep research with interactive visualizations, widgets and dashboards, and motion design and video editing, powered by its native multimodal architecture.
Native Multimodality & Long Context : Kimi K3 understands text, images, and video within the same model, and supports a 1-million-token context window.
Open Frontier Weights : We release the full Kimi K3 model weights under the Kimi K3 License, making frontier intelligence openly available for research, deployment, and further innovation.
> Currently Kimi K3 requires a Pro or Max subscription, and **consumes extra usage credits**. We’re quickly working on adding capacity to expand access.
Kimi K3 is an open-weight, native multimodal agentic model and our most capable model to date. It is a 2.8T-parameter model built on Kimi Delta Attention (KDA) and Attention Residuals (AttnRes), with native vision capabilities and a 1-million-token context window. It is the world's first open 3T-class model, designed for frontier intelligence across long-horizon coding, knowledge work, and reasoning.
### Key Features<br>- **New Architecture**: Kimi K3 is built on Kimi Delta Attention (KDA) and Attention Residuals (AttnRes), and scales up MoE sparsity with a Stable LatentMoE framework that activates 16 out of 896 experts — yielding an approximate 2.5× improvement in overall scaling efficiency over Kimi K2.<br>- **Long-Horizon Coding**: Operating with minimal human oversight, Kimi K3 sustains long engineering sessions, navigates massive repositories, and orchestrates terminal tools — from GPU kernel optimization and compiler development to vision-in-the-loop game dev, CAD, and even chip design.<br>- **Agentic Knowledge Work**: Kimi K3 advances end-to-end knowledge work, producing deep research with interactive visualizations, widgets and dashboards, and motion design and video editing, powered by its native multimodal architecture.<br>- **Native Multimodality & Long Context**: Kimi K3 understands text, images, and video within the same model, and supports a 1-million-token context window.<br>- **Open Frontier Weights**: We release the full Kimi K3 model weights under the Kimi K3 License, making frontier intelligence openly available for research, deployment, and further innovation.
Paste, drop or click to upload images (.png, .jpeg, .jpg, .svg, .gif)