Laya eats Jev’s lunch

Laya eats Jev’s lunch.

It is really wild how fast tech is moving right now.

Recently I talked about Jev and how it is actually a somewhat significant innovation and not just another version of an LLM.

Now, only a week later, Laya gets put on my radar.

Laya is an Open-source version of Jev that can be run even cheaper and possibly faster.

Again, this is not your LLM spitting out tokens mainly trying to mimic conversations.

This behaves more like a classifier, but a general one. You feed it a question, and it spits back answers with probability scores.

So instead of getting a slow, bloated response from an overpriced LLM, you can get lightning-fast, actionable responses from these decision-focused models.

What is the trade-off? From what I am seeing, Jev still beats Laya in accuracy, so for your more complex classifications, you still might want to stick with Jev.

These innovations are amazing, and the fact that it all seems to be a race to the bottom price-wise is great for those of us building on these innovations.

Just make note of that if you are an investor in these companies.

Have any of you played with these new styles of models? What are you building with them?