The easiest entry point for Kimi K3.1 Pro is Kimi's website and mobile app. After logging in with a free account, it is the default model in the input box, with thinking intensity set to High — just send a question and use it. Writing code goes through the terminal, and products and scripts go through the API. The entry points, costs, and account requirements of the three paths are completely different, so tell them apart before you start, and do not use the chat method to call the API.
What each of the three paths requires
The chat path only needs a free account, and is suited to trying prompts, reading documents, and looking at images. The official side has not announced a message quota for the free plan, so do not treat it as a production channel. The API path requires topping up at least 1 dollar first to unlock it. The model name is kimi-k3.1-pro, input costs 2 dollars per million tokens, output costs 10 dollars per million tokens, the interface is compatible with the OpenAI format, the base address is Moonshot's API address, and the key is created in the official console. The terminal path uses Kimi Code. Logging in uses an account, with no need to manage a key yourself, but it is a paid membership benefit: the free Adagio plan does not include it, only Plus members and above can use it, and the Enterprise edition also requires an administrator to enable it manually.
When something goes wrong, first tell which path you are on
If you cannot find the model in chat, first confirm whether you are logged in with the same account; an Enterprise account needs an administrator to confirm whether it has been enabled. For API errors, check two things first: whether you have successfully topped up to unlock it, and whether the key and model name are written correctly — insufficient balance will reject the request outright. For terminal login failure, first confirm whether your membership level is high enough, then check whether the login command is from the current version. There is one more boundary to remember: Kimi K3.1 Pro was released on 23 October 2026, and its context window is 1048576 tokens. Long documents fit, but input tokens are still billed, so split large files first and feed only the parts you need, or the bill will get out of control.
