Baby Potato: our most compact model yet
There’s a third model in the Potato AI line-up: Baby Potato. It’s a small language model for anyone who doesn’t need a whole Village Potato.
Why use a small language model?
Large models can do a lot, but they need a lot of power and a lot of hardware. A small language model runs right on the device — that’s what people mean by edge AI. Your prompt never travels to the cloud, so nobody can intercept it or keep it.
Baby Potato takes the idea as far as it goes: the device is the model. It fits in the palm of your hand or, if you’re feeling ambitious, on the tip of your finger.
How the Potato AI models compare
- Village Potato — large and contemplative, for complex tasks.
- Central Asian Potato — a balance of reasoning and speed.
- Baby Potato — small and nimble, for quick questions.
All three models are equally hallucination-free. The difference lies in how much energy they build up before answering and how long they keep in the cellar.
Who Baby Potato is for
Anyone who wants to try a local AI model without buying a GPU, a cluster or a separate computer. Baby Potato comes with the Young plan and works from day one after delivery.
Small model, small carbon footprint. Baby Potato’s is negative.
Limitations
Baby Potato is faster than its older siblings, but “faster” is a relative term here. If you need an answer by morning, ask the night before.
