Free & occasionally unstable AI model gateway — you get what you pay for.
Point any OpenAI-compatible client at the base URL and go. No SDK, no signup, no mercy.
OpenAI-compatible chat completions with streaming support on most models.
curl http://jankrouter.waifly.com/v1/chat/completions \ -H "Content-Type: application/json" \ -d '{ "model": "qwen3-guard-8b", "messages": [ {"role": "user", "content": "Hi"} ] }'
Text-to-image generation across 6 models, from SDXL Lightning to Nanobanana Pro.
curl http://jankrouter.waifly.com/v1/images/generations \ -H "Content-Type: application/json" \ -d '{ "model": "lucid-origin", "prompt": "A cat astronaut in space", "size": "1024x1024" }'
Edit and inpaint images across 5 supported models.
curl http://jankrouter.waifly.com/v1/images/edits \ -H "Content-Type: application/json" \ -d '{ "model": "lucid-origin", "prompt": "Change the cat to a dog", "image": "base64data...", "size": "1024x1024" }'
Zero-effort one-liner for humans. Paste a URL, get text back.
http://jankrouter.waifly.com/generate/Hello?model=qwen3-guard-8b
Some models require a Pro API Key created by Marcos F R Games. Pass it via the Authorization: Bearer <token> header or directly in the pro parameter.
A few models don't support streaming. If you request streaming on those, you'll receive an error instead of magic.
Every model, every alias. Click an alias to copy it. Badges tell you what you're getting into.
no models matched. even the router is confused.