Kimi K3 is Kimi's most powerful flagship model with 2.8 trillion parameters. Built on KDA hybrid linear attention (Kimi Delta Attention) and Attention Residuals technologies, it natively supports vision understanding and features a 1 million token context window. It is the world's first open-source 3-trillion-parameter model designed for long-context programming, knowledge work, and advanced reasoning.
Inference Service Provider
The inference service provider for kimi/kimi-k3 is Moonshot AI.
Model Capabilities
| Capability | Support | Capability | Support |
|---|---|---|---|
Input Modality |
Text Image Video |
Output Modality |
Text |
Model Experience |
Function Calling |
||
Structured Outputs |
Web Search |
||
Prefix Completion |
Context Caching |
||
Batch Inference |
Fine-tuning |
Context Limits
| Parameter | Value | Parameter | Value |
|---|---|---|---|
Max Input Length |
1048576 |
Max Output Length |
1048576 |
Context Window |
1048576 |
Max Input Length (Thinking Mode) |
1048576 |
Max Output Length (Thinking Mode) |
1048576 |
Max Chain-of-Thought Length |
1048576 |
Pricing
This page only shows the original pricing for model API calls, excluding any limited-time promotions. Visit Model Studio Console for promotional offers.
China (Beijing)
| Billing Item | Price (CNY) | Unit |
|---|---|---|
Input |
20 |
Per 1M tokens |
Output |
100 |
Per 1M tokens |
Input(Implicit Cache) |
2 |
Per 1M tokens |
Rate Limits
China (Beijing)
| Parameter | Value |
|---|---|
RPM (Requests Per Minute) |
500 |
TPM (Tokens Per Minute) |
3,000,000 |