Kimi K3 is Moonshot AI's open-weight mixture-of-experts model, announced in July 2026 with 2.8 trillion total parameters. Moonshot built it for long-horizon coding, knowledge work and reasoning, and says it can sustain long engineering sessions, move around very large repositories and drive terminal tools. It also understands images and video natively. In Jantaris it has a context window of 1,048,576 tokens and up to 131,072 output tokens. Moonshot itself notes that K3 still trails the strongest proprietary models overall.
Best for
- Long coding sessions in big repositories
- Terminal-driven agents that run build and test loops
- Research and knowledge work over long inputs
- Teams that want an open-weight model they could later host themselves
How to call it
Use the OpenAI SDK and change the base URL:
from openai import OpenAI
client = OpenAI(base_url="https://jantaris.com/v1", api_key="sk-jt-...")
resp = client.chat.completions.create(
model="kimi-k3",
messages=[{"role": "user", "content": "Find why the build fails and propose a fix."}],
)
print(resp.choices[0].message.content)FAQ
What is the Kimi K3 API price?
$4.05 / $20.25 per 1M tokens in Jantaris. You top up once and the same balance covers Kimi K3 and every other model, with no Moonshot account needed.
What are Kimi K3's known weaknesses?
Moonshot lists them openly. Agent tools that cut or rewrite the model's chain of thought can lower quality. In ambiguous situations K3 tends to act instead of asking for clarification, so give precise instructions. Moonshot also says its conversational polish trails the top closed models.
How does Kimi K3 compare with Claude for coding?
Moonshot positions K3 as a strong open model that is usually within reach of the frontier but not ahead of it. The practical test is your own repository: run the same task on K3 and Claude Sonnet 5 in Jantaris, then compare the diffs and the cost of each run.
How to call Kimi K3 via API
from openai import OpenAI
client = OpenAI(base_url="https://jantaris.com/v1", api_key="sk-jt-...")
resp = client.chat.completions.create(
model="kimi-k3",
messages=[{"role": "user", "content": "Write a haiku about Baltic amber."}],
)
print(resp.choices[0].message.content)curl https://jantaris.com/v1/chat/completions \
-H "Authorization: Bearer $JANTARIS_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "kimi-k3", "messages": [{"role": "user", "content": "Write a haiku about Baltic amber."}]}'import OpenAI from "openai";
const client = new OpenAI({ baseURL: "https://jantaris.com/v1", apiKey: process.env.JANTARIS_API_KEY });
const resp = await client.chat.completions.create({
model: "kimi-k3",
messages: [{ role: "user", content: "Write a haiku about Baltic amber." }],
});
console.log(resp.choices[0].message.content);