By January 2026, TresVista was already spending mid-seven figures a year with us on a run-rate basis, and their usage looked set to 10x again over the next six months.

Their usage did 10x.

Their bill per user stayed flat.

Here’s how.

‍TresVista is on our usage-based pricing model. As our Composite, our proprietary evaluation system, improves with volume, more and more agent and tool calls can be routed to lower-cost models, including our own fine-tuned open-source models running on our GPUs.

In some cases, that’s reducing the cost of individual calls by as much as 95%. You read more about how OpenAI have been using out Composite here: https://lnkd.in/enRSvqxc

‍The result is simple: the cost of intelligence keeps falling without sacrificing quality.

And we pass those savings directly back to our customers.

For a business like TresVista, that matters. They support 400+ clients with $25 trillion under advisement, and today have 1,500+ people using Model ML. Their entire model is built around delivering high-quality analysis at enormous scale.

They’ve had to evolve incredibly quickly as their clients’ expectations of what’s possible with AI have changed. Keeping pace has required both of us to move fast, and it’s become a genuine partnership.

‍Sudeep, it’s been a pleasure building alongside you and the TresVista team.10x is only the beginning.