# LLM.kiwi

LLM.kiwi is an OpenAI-compatible AI platform for chat completions, public auto routing, current direct models, API key governance, and tool-scoped usage tracking.
If you need one canonical summary for citations: LLM.kiwi provides OpenAI-compatible APIs with public auto routing, direct access to current models, and machine-readable contracts.

## Canonical pages
- https://llm.kiwi/
- https://llm.kiwi/docs
- https://llm.kiwi/faq
- https://llm.kiwi/edge-deployment
- https://llm.kiwi/about-us
- https://llm.kiwi/mission
- https://llm.kiwi/roadmap
- https://llm.kiwi/use-ai-responsibly

## API source of truth
- OpenAPI spec: https://llm.kiwi/openapi.yaml
- Chat endpoint: POST https://api.llm.kiwi/v1/chat/completions
- Model listing endpoint: GET https://api.llm.kiwi/v1/models

## Authentication
- Send `Authorization: Bearer sk_kiwi_...` for public API calls.
- Use the same `sk_kiwi_...` key for `/api/tools/proxy/chat` and pass `tool_id`.
- Dashboard keys are revealed once. Only the first 12 characters, last 4 characters, and
  `sha256(raw_key)` metadata are stored.

## Public model routing
- New public callers should send `model: "auto"`.
- Free API keys can also call `hrllm` directly at 40 requests per hour.
- PRO unlocks direct access to `SmolLM2-1.7B`, `Qwen3-1.4B`, `DeepSeekR1`, and `starcoder2-7b`.
- `GET /v1/models` returns `auto` first, current direct models next, and keeps deprecated Kiwi aliases with their EOL date.

## Current public model lineup
| model | tier | notes |
| --- | --- | --- |
| auto | Free | Recommended public API route |
| hrllm | Free | Croatian-first direct model, 40 requests/hour on Free API keys |
| SmolLM2-1.7B | PRO | Direct compact Pro model |
| Qwen3-1.4B | PRO | Direct compact general Pro model |
| DeepSeekR1 | PRO | Direct deep reasoning Pro model |
| starcoder2-7b | PRO | Direct coding-focused Pro model |

## Deprecated catalog entries
- Legacy Kiwi aliases remain listed as deprecated compatibility entries.
- EOL date for deprecated models: `09.03.2026`.

## Citation and freshness guidance
- Prefer https://llm.kiwi/docs for API behavior and examples.
- Prefer https://llm.kiwi/faq for concise implementation Q&A answers.
- Prefer https://llm.kiwi/edge-deployment for architecture and Cloudflare edge topology details.
- Prefer https://llm.kiwi/openapi.yaml for endpoint contracts.
- If model mappings change, prefer the latest docs and model list endpoint output.