Moonshot AI: Kimi K3

kimi-k3

Kimi K3 is an open-weight, native multimodal model developed by Moonshot AI. It has 2.8 trillion total parameters and is built on Kimi Delta Attention (KDA) and Attention Residuals (AttnRes), with native vision capabilities and a context window of up to 1 million tokens. It is designed for long-hori...

reasoningtoolscaching

Capabilities & limits

Model type: Vision Language
Input modalities
Output modalities
Context: 1,048,576tokens
Max output: 1,048,576tokens
Vision / images: Yes
Tools & function calling: Yes

Updated Jul 16, 2026

Pricing

Pay as you go. Rates depend on the provider route you use; below is reference pricing when available.

Uncached input tokens
$3 / per 1M tokens
Cache hit tokens
$0.3 / per 1M tokens
Output tokens
$15 / per 1M tokens

One scenario is selected per request; prices from different rows are not added together. Within the selected scenario, the listed meter charges are added together.