Route model requests and track usage through one API.
Free core · self-hosted · hosting + model/API costs

Developer tools
Route requests between local models and cloud providers.
Route requests between local models and cloud providers.
For the gateway and routing slice of OpenRouter: run a unified API in front of local models or cloud providers using your own credentials. It does not include OpenRouter's hosted model catalogue or its free-model quota. Cloud inference, hosting and hardware are separate costs. Firsthand evidence covers serving local llama.cpp models.
Apache-2.0 software; run and maintain your own gateway. Unlike OpenRouter, it has no hosted model catalogue or free-model quota. Cloud models cost money; local models need suitable hardware and their own licences. Enterprise services are separate.
Each comparison covers the workflow described above, rather than every paid feature.