Liquid d1-3B

The claim

An open-weight text-and-image decision model for local deployment. It offers a way to compare hosted inference with hardware you control.

Our read

The weights use a custom license, not Apache or MIT. Review its terms for your intended use. Published latency depends on hardware and input conditions. Use the documented decision runtime rather than assuming a generic generation example serves this checkpoint correctly.

Source reviewed: Primary sources support the stated scope. Not installed or inference-tested by Typed Recipes.

Test status

We haven't reproduced this yet. When we do, measured accuracy, latency, and cost will appear here next to the claimed figures.

Related entries