Skip to main content
Models/Moonshot AI/Kimi K3

Kimi K3

Moonshot AI 🧠 Language Model

Moonshot's open-weight 2.8T-parameter multimodal reasoning model with a 1M-token context window.

Specifications

Input modalitiestext, image
Output modalitiestext
Capabilitieschat, reasoning, function_calling, agentic
Relative speed●●●○○ Average
Context window1,048,576 tokens
Max output131,072 tokens
Vision inputYes
Tool callingYes
Streaming outputYes
Input price$5.25 / 1M
Output price$26.25 / 1M

Pricing

$5.25 / 1M input tokens  ·  $26.25 / 1M output tokens

Platform credit pricing — live rates on the Models page.

Use Kimi K3 on Zubnet →

More from Moonshot AI

K
Kimi K2.6
Latest Kimi K2 reasoning model. Replaces K2 thinking variants (k2-thinking and k2-thinking-turbo) being retired 2026-05-25.
🧠 Language Model
K
Kimi K2.7 Code
Coding-specialized Kimi K2 model (k2.7). 256K context, agentic tool use and function calling, tuned for software engineering.
🧠 Language Model
K
Kimi K2.7 Code Highspeed
High-speed variant of Kimi K2.7 Code for low-latency agentic software engineering.
🧠 Language Model
K
Kimi K2 Thinking
Most powerful open-source thinking model. 1T MoE with 32B active, 256K context, native INT4 quantization.
🧠 Language Model