moonshot/kimi-k3
Kimi K3 is Moonshot AI's open-weight flagship: a 2.8T-parameter sparse-MoE model (104B active) with a 1M-token context window and native image understanding. Reasoning is always on, with three effort levels (low / high / max) selected via reasoning_effort, and thinking is streamed as separate reasoning_content deltas. Supports tool calling, JSON Schema structured output and automatic prefix caching. Available through the OpenAI-compatible and Anthropic Messages interfaces.
Source: official_x0.5 · Verified 2026-09-10
Cache read $0.15/M · Cache write —/M
Use either ID to call this model via the API.
moonshot/kimi-k3Replace the ONEHOP_KEY placeholder with your API key. Create one →
from openai import OpenAI
client = OpenAI(
base_url="https://api.onehop.ai/v1",
api_key="<ONEHOP_KEY>",
)
completion = client.chat.completions.create(
model="moonshot/kimi-k3",
messages=[{"role": "user", "content": "What is the meaning of life?"}],
)
print(completion.choices[0].message.content)