Is LLM.kiwi really free?
Is LLM.kiwi really free?
LLM.kiwi is currently free and does not require a credit card. Free access is still subject to capacity, rate limits, and fair use. See Limits and pricing.
Do I need to download or host a model?
Do I need to download or host a model?
No. Your application sends HTTPS requests to the hosted LLM.kiwi API. You do not need a GPU, Docker, or a model server.
How do I get an API key?
How do I get an API key?
Sign in at llm.kiwi/login, open the dashboard, and create a key. Copy it when shown and store it safely.
What is the base URL?
What is the base URL?
Use
https://api.llm.kiwi/v1. In an SDK, set this as the base URL. In direct HTTP, append an endpoint such as /chat/completions.Which model should I choose?
Which model should I choose?
Start with
auto for general use. Test hrllm when Croatian-language quality is your main requirement. Query GET /v1/models for current availability.Is the API OpenAI-compatible?
Is the API OpenAI-compatible?
LLM.kiwi supports the OpenAI-compatible chat completions patterns documented here. Set the custom base URL in a compatible client. Features not listed in the LLM.kiwi API reference should not be assumed to work.
Can I use it from React or browser JavaScript?
Can I use it from React or browser JavaScript?
Your frontend can use an AI feature, but it should call your backend. The backend holds the API key and calls LLM.kiwi. Never embed a private key in code sent to the browser.
Can I use it from a mobile app?
Can I use it from a mobile app?
Use your own backend for the same reason as a browser app: secrets packaged with a mobile application can be extracted.
Does the API remember conversations?
Does the API remember conversations?
No. Include the relevant system, user, and assistant messages with each request. Your application decides what history to store and resend.
Why does the same prompt produce different answers?
Why does the same prompt produce different answers?
Language-model output can vary because of sampling and model routing. If consistency matters, make instructions precise, test model parameters supported by the endpoint, and validate the output.
Can I require JSON output?
Can I require JSON output?
You can request JSON, but your code must still parse and validate it. Define a schema, reject invalid fields, and provide a fallback when parsing fails.
What should I do if a key leaks?
What should I do if a key leaks?
Revoke it immediately in the dashboard, create a replacement, update the application’s secret, and remove the exposed value from code and logs. Deleting only the latest Git commit may not remove it from history.
Why am I receiving a 401 error?
Why am I receiving a 401 error?
The key is missing, malformed, invalid, or revoked. Verify
Authorization: Bearer sk_kiwi_... and confirm that the environment variable is available to the running process.Why am I receiving a 429 error?
Why am I receiving a 429 error?
A request or capacity limit was reached. Wait, reduce request frequency or concurrency, and retry a limited number of times with exponential backoff.
Can I use LLM.kiwi in production?
Can I use LLM.kiwi in production?
Evaluate it against your reliability, privacy, capacity, and support requirements. At minimum, keep keys server-side, set timeouts, handle failures, validate model output, and provide a fallback. See Best practices.
Where can I get help?
Where can I get help?
Follow Troubleshooting first. If the problem remains, email support@llm.kiwi with the status code, approximate time, SDK version, and a minimal redacted request. Never send a full API key.