It is coming.

The most requested feature since we open sourced oneinfer-edge.

We are adding a "Quantization Playbook" as a standalone feature inside the oneinfer-edge open source repo.

Right now, quantizing a model for local deployment means researching the right format for your hardware, running separate tools, debugging format incompatibilities with your serving library, and then starting the deployment process from scratch. A multi-hour process that sits completely outside your inference control plane.

With the Quantization Playbook inside oneinfer-edge - your hardware is already scanned. Your serving library is already detected. Your model is already known. Quantization becomes the natural step before deployment - not a separate workflow you manage outside the app.

Pick your model. oneinfer-edge handles the quantization. Deploy directly from the same control plane.

And before you deploy - you see exactly what you are trading. Perplexity. Token accuracy. Quality delta across quantization levels. So the decision is not a guess. It is a tradeoff you make with full visibility.

This is part of our commitment to making oneinfer-edge the only open source tool a developer needs to go from raw model to running inference - local, cloud, or both.

More details dropping soon.

Repo in the comments below.

Star the repo. You will not want to miss this one.