About Kimi K3
Kimi K3 is Moonshot AI's largest Mixture-of-Experts model with a 1M-token context window, built for long-horizon agentic coding and reasoning workloads with native visual understanding.
The model uses the Kimi Delta Attention (KDA) hybrid linear attention mechanism and activates 16 of 896 experts per token. Thinking mode is always enabled, with reasoning content returned alongside the answer. It handles tool calling with strict JSON schema output, and understands images and video. Open weights were released in July 2026 under the custom Kimi K3 License.
| Metric | Value |
|---|---|
| Parameter Count | 2.8 trillion |
| Mixture of Experts | Yes (16 of 896 experts active) |
| Context Length | 1,048,576 tokens |
| Multilingual | Yes |
| Quantized* | Unknown |
*Quantization is specific to the inference provider and the model may be offered with different quantization levels by other providers.
Ready to build with Kimi K3?
Try Kimi K3 in the Workbench to prompt it, compare outputs, and iterate on prompts without writing any code. When you're ready to ship, call the same model from our API and build your own apps on top of it.
curl -sSf -X POST https://hub.oxen.ai/api/ai/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $OXEN_API_KEY" \
-d '{
"model": "kimi-k3",
"messages": [
{
"role": "user",
"content": "Try sending a message."
}
]
}'