# RainDragon AI

> RainDragon AI is an LLM inference API serving open-weight models (Google Gemma 4 26B-A4B MoE and Gemma 4 31B for text, OpenAI Whisper for speech-to-text) at https://inference.raindragon.ai. It is OpenAI-compatible (base URL https://inference.raindragon.ai/v1: chat completions, Responses API, models, files and batches, audio transcriptions) and Anthropic-compatible (POST /v1/messages at base URL https://inference.raindragon.ai). Pricing is pay per token with no subscription or minimum spend.

- Authenticate with `Authorization: Bearer <API key>`; create keys in the dashboard at https://inference.raindragon.ai/dashboard.
- Model ids: `google/gemma-4-26B-A4B-it`, `google/gemma-4-31B-it`; speech-to-text: `openai/whisper-large-v3`, `openai/whisper-large-v3-turbo`, `distil-whisper/distil-large-v3`.
- Live model list, per-token prices and context lengths: `GET https://inference.raindragon.ai/api/models` (JSON, no key needed) and `GET https://inference.raindragon.ai/v1/models`.
- Coding agents (OpenCode, Cline, Aider, Codex CLI, Claude Code, ...) work by pointing their OpenAI or Anthropic base URL at RainDragon; see the coding tools page.

## Docs

- [API Documentation](https://inference.raindragon.ai/api-documentation.md): RainDragon AI API reference: OpenAI-compatible chat completions, Responses and Anthropic Messages APIs, models, authentication, reasoning, tools, prompt caching, batch, flex, fill-or-kill, rate limits and errors.
- [Coding Tools Setup](https://inference.raindragon.ai/coding-tools.md): Use RainDragon AI models in coding agents: base URL and API key setup for OpenCode, Cline, OpenHands, Aider, Codex CLI, Claude Code, Qwen Code and more, with measured speeds.
- [AI Gateways](https://inference.raindragon.ai/gateways.md): Route RainDragon AI requests through Cloudflare AI Gateway, Portkey, Helicone or Bifrost as a custom OpenAI-compatible provider.
- [Management API](https://inference.raindragon.ai/management-api.md): RainDragon AI Management API: manage fallback providers and read usage metrics and engine status with a scoped management key.
- [Pricing](https://inference.raindragon.ai/pricing.md): RainDragon AI pricing: pay per token with no subscription or minimum spend. Live per-model prices are served as JSON at /api/models.

## API reference

- [OpenAPI spec](https://inference.raindragon.ai/openapi.json): OpenAPI 3.1 description of the public endpoints (chat completions, Responses, Anthropic Messages, models, files and batches, audio transcriptions, Management API).
- [Models and prices](https://inference.raindragon.ai/api/models): live JSON list of models with per-token prices, context lengths and billing rules.

## Optional

- [Full docs as one file](https://inference.raindragon.ai/llms-full.txt): every page above concatenated as Markdown.