Models & pricing / kimi-k3 · Moonshot
Moonshot

kimi-k3 API

Kimi K3 is Moonshot’s previous-generation flagship for long-horizon coding, knowledge work and visual understanding, with its own kimi-k3 API identifier. On XiuAI, compare it with K3.1 Pro using your integration, tool loop and current pricing.[1][2]

kimi-k3

Model overview

Published provider specifications. Available features depend on the current API route.

Input
text / image / video
Output
Text
Context
1M
API routes
OpenAI-compatible API

Existing projects and visual feedback

The documented use cases include long coding tasks and frontend, game or CAD work guided by screenshots. Supply repository constraints, tool permissions and acceptance criteria. The model produces text or tool requests; your application supplies file operations and the execution environment.[1]

The vendor now calls K3 a previous-generation flagship. Keep a baseline for an existing application and evaluate K3.1 Pro on the same tasks before migrating. “Previous generation” does not mean it is delisted on XiuAI, nor does a newer name determine quality or cost.[1][2]

Thinking and tool history

K3 documentation states that thinking stays on, with top-level reasoning_effort set to low, high or max. In multi-turn and tool workflows, append the full returned assistant message rather than retaining only content. Handle streamed reasoning separately from the final answer.[1]

Comparing a complete tool-loop cost

Multi-turn input, tool results and output all affect a long task’s total usage. Compare the complete task using current rates and request records. Kimi’s direct-service recharge thresholds, rate tiers and coupons do not define an XiuAI account.[1][2]

API access

Using K3 through XiuAI

Use kimi-k3, your XiuAI key and balance, and the OpenAI-compatible endpoint in the current catalog. Configure the Base URL and authentication for XiuAI; do not mix an XiuAI key with the vendor domain or MOONSHOT_API_KEY setup from a direct-service example.[1][2]

Send a request using your existing API key.

POST /v1/chat/completions
curl 'https://router-api.xiu.ai/v1/chat/completions' \
  -H "Authorization: Bearer $XIUAI_API_KEY" \
  -H 'Content-Type: application/json' \
  --data-raw '{
  "model": "kimi-k3",
  "messages": [
    {
      "role": "user",
      "content": "Hello"
    }
  ]
}'
Read the API documentation

Frequently asked questions

Is K3.1 Pro an update behind the same K3 API name?

They are separate catalog entries: kimi-k3 and kimi-k3.1-pro are distinct IDs. Check price, parameters and actual output when switching instead of assuming the new model inherits all behavior.[2][1]

Why should I keep more than the previous text answer?

The vendor requires the full assistant message for multi-turn and tool calls. Keeping only content can discard tool or reasoning fields. Preserve protocol state instead of treating displayed text as the complete conversation.[1]

Does screenshot understanding mean it operates CAD or a terminal for me?

No. Vision helps analyze input; tool operation requires an environment connected and authorized by your application. A complete vendor demonstration does not establish an equivalent XiuAI workspace.[1]

Related models

kimi-k3.1-pro

Sources and updates

Specifications reference provider documentation; prices and availability come from the current XiuAI catalog. Specifications checked: 2026-11-11

Content reviewed: 2026-11-16

[1] Kimi K3 documentation[2] Current XiuAI model and pricing data