Liquid AI's 3B classifier answers in 16 milliseconds on a Jetson, and its weights are downloadable
A decision model does not generate tokens. It takes a state and a question and returns an answer in a single forward pass, which is why the latency figures look like this.