Alex Atallah

like optimizing your Datadog bill.

I think the future is going to be multi-model for many reasons, and that's a big one today, and it's a big one this coming year.

There are a lot of companies who are also just doing one task that should be broken up into multiple tasks.

Yes.

served by specific lower-cost models, and they're getting massive cost savings out of it in addition to improving their recall and accuracy.

So it's really important to do.

And then down the road, another use case will be getting better than state-of-the-art performance by using multiple models.

And that's also helping our customers.

orchestrating multiple models that were trained by completely different companies to do something for you and then using a judge or some other set of heuristics to select the best result or combine them together.

Good question.

A lot of people think that basically everybody's just using one model, and that's completely untrue.

We have about 350 models that are being used by hundreds of active users per day.

And while...

And the token diversity is growing over time.

If you go to our rankings page, we have a chart that just shows the graph getting more and more diverse.

We're doing about 120 trillion tokens per month now.

So there's diversity on model selection, but there's also a ton of diversity in providers for specific models.

It used to be the case that being a provider is just really hard because how do you get distribution?

Even if you decide to specialize in very, very, very low-cost but slow inference, so stuff that's optimized for batch agentic workloads, how are people going to discover it?

We're building all these SKUs in the marketplace so that providers can find a market really, really quickly.