Question 1

What exactly do I get when I buy a package?

Accepted Answer

A self-contained kit: the optimized model weights, the exact serving configuration, and the measured benchmark receipt. You download it after purchase and it is yours forever - nothing calls home, nothing expires.

Question 2

Where can I run it?

Accepted Answer

On any cloud with the package's GPU class, or on your own hardware. Deploy targets like Docker, Compose, and Kubernetes ship inside the kit.

Question 3

Are the performance numbers real?

Accepted Answer

Yes. Every number is measured on the named GPU against the unoptimized baseline of the same model. The full methodology and receipt ship inside the kit so you can reproduce it.

Question 4

How is this different from a metered API?

Accepted Answer

A metered API charges per token forever and keeps the model behind its wall. Here you buy once, the weights run on your infrastructure, your prompts never leave your network, and your cost is your GPU.

Question 5

What about the model license?

Accepted Answer

Each package names its upstream license from the model author, and the full license text ships inside the kit.

Optimized models, with the proof attached

Common questions

Deploy your first optimized model, measured before you ship